Plant intelligence based metaheuristic optimization algorithms
Abstract
Classical optimization algorithms are insufficient in large scale combinatorial problems and in nonlinear problems. Hence, metaheuristic optimization algorithms have been proposed. General purpose metaheuristic methods are evaluated in nine different groups: biologybased, physicsbased, socialbased, musicbased, chemicalbased, sportbased, mathematicsbased, swarmbased, and hybrid methods which are combinations of these. Studies on plants in recent years have showed that plants exhibit intelligent behaviors. Accordingly, it is thought that plants have nervous system. In this work, all of the algorithms and applications about plant intelligence have been firstly collected and searched. Information is given about plant intelligence algorithms such as Flower Pollination Algorithm, Invasive Weed Optimization, Paddy Field Algorithm, Root Mass Optimization Algorithm, Artificial Plant Optimization Algorithm, Sapling Growing up Algorithm, Photosynthetic Algorithm, Plant Growth Optimization, Root Growth Algorithm, Strawberry Algorithm as Plant Propagation Algorithm, Runner Root Algorithm, Path Planning Algorithm, and Rooted Tree Optimization.
Keywords
Plant intelligence Global optimization Metaheuristic methods1 Introduction
Most of the optimization algorithms need mathematical models for system modelling and objective function. Establishment of a mathematical model is mostly hard for complex systems. Even though the model is established, it cannot be used due to high cost of solution time. Classical optimization algorithms are insufficient in largescale combinational and nonlinear problems. Such algorithms are not effective in adaptation a solution algorithm for a given problem. In most cases, this requires some assumptions whose approval of validity can be difficult. Generally, because of natural solution mechanism of classical algorithms, significant problem is modeled the way that algorithm can handle it. Solution strategy of classical optimization algorithms generally depends on type of objective and constraints (linear, nonlinear etc.) and depends on type of variables (integer, real) used in modelling of problem. Furthermore, effectiveness of the classical algorithms highly depends on solution space (convex, nonconvex etc.), the number of decision variable, and the number of constraints in problem modelling.
Another important deficiency is that; if there are different types of decision variables, objectives, and constraints, general solution strategy is not presented for implemented problem formulation. In other words, most of algorithms solve models which have certain type of object function or constraints. However, optimization problems in many different areas such as management science, computer, and engineering require concurrently different types of decision variables, object function, and constraints in their formulation. Therefore, metaheuristic optimization algorithms are proposed. These algorithms become quite popular methods in recent years, due to their good computation power and easy conversion. In other words, a metaheuristic program, written for a specific problem with a single objective function, can be adapted easily to a multi objective version of this problem or a different problem.
Studies in recent years indicate that plants also exhibit intelligent behaviors. According to this, it is thought that plants have nervous system. For example, in roots, imported light and poison information are transmitted to growth center in root tips and the roots perform orientation according to that. Additionally, it is thought that plants make contact with external world using electric current. An example for this is defense mechanism which is shown against to aphid or caterpillar by plants. After the first developed attack, plants produce secretions which can make worse taste or which can poison their enemy.
Inspiring by the flow pollination process of flowery plants, Flower Pollination Algorithm (FPA), has been developed in 2012 by Yang (2012). Invasive Weed Optimization (IWO), inspired from the phenomenon of colonization of invasive weeds in nature, has been proposed by Mehrabian and Lucas (2006). It is based on ecology and weed biology. Paddy Field Algorithm (PFA), which simulates growth process of paddy fields, is another plant based metaheuristic algorithm and has been proposed by Premaratne et al. (2009). Root Mass Optimization (RMO) is based on the process of root growth and has been proposed by Qi et al. 2013. Artificial Plant Optimization Algorithm (APOA) simulates growth model of plants which includes photosynthesis and phototropism mechanism (Zhao et al. 2011). Sapling Growing up Algorithm (SGuA) is a computational method based on cultivating, growing up, and mating of saplings and it is developed for efficient solutions to search and optimization problems (Karci 2007a).
Photosynthetic Algorithm (PA) is based on the processes of Calvin–Benson cycle and photorespiration cycle for the plants (Murase 2000). Plant Growth Optimization (PGO) is proposed to simulate plant growth with a realistic way considering the spatial occupancy, account branching, phototropism, and leaf growth. The main purpose of the model is to select active point by comparing the morphogen concentration for increasing the Lsystem (Cai et al. 2008). Root Growth Algorithm (RGA) is another plant based metaheuristic algorithm and simulates root growth of plants based on Lsystem (Zhang et al. 2014).
Strawberry algorithm has been proposed as an exemplar of Plant Propagation Algorithm (PPA). It maps an optimization problem onto survival optimization problem of strawberry plant and adopts strategy of survival in the environment for searching points in the search space which give the best values and ultimately the best value (Salhi and Fraga 2011). Runner Root Algorithm (RRA) is inspired by plants such as strawberry and spider plants which are spread through their runners and also which develop roots and root hairs for local search for minerals and water resources (MerrikhBayat 2015). Path Planning Algorithm is based on plant growth mechanism (PGPP) which uses phototropism, negative geotropism, apical dominance, and branching as basic rules (Zhou et al. 2016). One of the most recent plant based algorithm is Rooted Tree Optimization (RTO) and it is based on the intelligent behaviors of roots which select their orientation according to the wetness (Labbi et al. 2016). Researches about new versions and applications of these algorithms incrementally continue.
In this work, in the second part, the information is given about metaheuristic optimization and it is mentioned that why metaheuristic algorithms are needed. In the third part, plant intelligence is mentioned; optimization algorithms developed inspired by plant intelligence are examined and related works with these algorithms are described. FPA, IWO, PFA, RMO Algorithm, APOA, SGuA, PA, PGO, RGA, Strawberry Algorithm as PPA, RRA, PGPP, and RTO are explained. In the fourth part, discussions about general evaluations of these algorithms are presented and in the fifth part, the work is concluded along with future works.
2 Metaheuristic optimization
In most of the real life problems, solution space of the problem is infinite or it is too large for assessment of all the solutions. Therefore, with evaluating solutions, a good solution is needed to be found in acceptable time. Actually, for such problems, evaluating solutions in acceptable time has the same meaning with evaluating “some solutions” in the whole solution space. Selection of some solutions depending on what and how they are selected changes according to metaheuristic method. It cannot be guaranteed that optimal solution is included in the solutions which get involved in the evaluation. Therefore, the solution, proposed by metaheuristic methods for an optimization problem, must be perceived as a good solution, not as an optimal solution (Cura 2008).
They are criteria or computer methods identified in order to decide effective ones of various alternative movements for achieving any purpose or the goal. Such algorithms have convergence property, but they cannot guarantee exact solution, they can only guarantee a solution which closes to exact solution.
 (a)
Optimization problem can have a structure that the process of finding the exact solution cannot be defined
 (b)
Metaheuristic algorithms can be much simpler from the point of decision maker, in terms of comprehensibility.
 (c)
Metaheuristic algorithms can be used as a part of process of finding the exact solution, and learning purpose.
 (d)
Generally, the most difficult parts of real world problems (which purposes and which restrictions must be used, which alternatives must be tested, how problems data must be collected) are neglected in the definitions made with mathematical formulas. Faultiness of the data used in process of determining model parameters can cause much larger errors than suboptimal solution produced by metaheuristic approach (Karaboğa 2011).
3 Plant intelligence optimization algorithms
As a result of previous studies, it is observed that plants have gender identity and immune system. Furthermore, recent studies show that plants exhibit intelligent behavior. According to this, it is thought that plants have nervous system. For example, in roots, imported light and poison data are transmitted to growth centers in root tips, and roots perform orientation according to this. As another consideration, plants are considered to be contacting with the external world by electric currents. Defense mechanism, which is shown against to aphid or caterpillar by plants, can be shown as an example for this. After the first attack takes place, plants will worsen their taste or produce secretions which can poison their enemy.
3.1 Flower Pollination Algorithm (FPA)
3.1.1 Characteristic of flower pollination
It is estimated that, in nature, there are over a quarter of a million types of flowery plants. Approximately 80 % of all plant species are flowery plant. Flowery plants have been evolving for more than 125 million years. The main objective of a plant is reproducing through pollination. About 90 % of pollens are transferred by biotic pollinators such as animals and insects and about 10 % of pollens are transferred by abiotic pollinators such as wind. Biotic pollinators visit only some types of flowers. Therefore, pollen transfer of same types of flowers will be maximum. This provides advantages for pollinators in terms of research cost and limited memory. Focusing on some unpredictable but potentially more rewarding new flower species can give better results (Yang et al. 2013).
Pollination can be achieved by crosspollination or selfpollination. Crosspollination, or hybridization can occur from a flowers plant of a different plant. Selfpollination occurs from the same flowers pollen or different flowers of the same plant. Biotic crosspollination can occur in longdistance, because of long distance flight of birds, bats, bees, etc. Thus, this is considered as global pollination. In addition to this, bees and birds are able to move with the Levy flight behavior (Yang et al. 2013).
3.1.2 FPA
 (1)
Biotic crosspollination can be thought as global pollination process. Pollencarrying pollinators move according to a Levy flight process (Rule 1).
 (2)
Selfpollination and local pollination are used for local pollination (Rule 2).
 (3)
Pollinators such as birds and bees can develop flower persistence (Rule 3).
 (4)
A switch probability \(p\in \left[ {0,1} \right] \), which is slightly biased towards local pollination, can control the switching or interaction of global and local pollination (Rule 4).
In principle, flower pollination activities can occur both at local and global levels, at all scales. However, substantially, the flowers in the notfaraway or adjacent flower patches are more likely to be pollinated by local flower pollen than those far away. Therefore, value of the switch probability can be taken as \(p=0.8\). The pseudo code of FPA is shown in Fig. 3.
3.1.3 Studies with FPA
Wang and Zhou (2014) proposed dimension by dimension improvement based FPA, for multiobjective optimization problem. They also applied local neighborhood search strategy in this improved algorithm for enhancing the local searching ability. Yang et al. (2013) indicated that, it is important to balance exploration and exploitation for any metaheuristic algorithm. Because, interaction of these two components could significantly affect the efficiency. Therefore, they studied FPA with Eagle Strategy.
AbdelRaouf et al. (2014a) used FPA with chaos theory for solving definite integral. Additionally, they proposed a hybrid method which was a combination of FPA and PSO for improving search accuracy. They used it for solving constrained optimization problems (AbdelRaouf et al. 2014b). They also proposed a new hybrid algorithm combined with FPA and chaotic harmony search algorithm, to solve Sudoku puzzles (AbdelRaouf et al. 2014c). Sundareswaran et al. proposed a modification for steps of traditional GA, used for PVM inverter, by imitating flower pollination and subsequent seed production in plants (Sulaiman et al. 2014). Prathiba et al. set the real power generations using FPA, for minimizing the fuel cost, in economic load dispatch, which is the main optimization task in power system operation (Prathiba et al. 2014).
Łukasik and Kowalski (2015) studied FPA for continuous optimization and they compared solutions with PSO. Platt used FPA in the calculation of dew point pressures of a system exhibiting double retrograde vaporization. The main idea was to apply a new algorithmic structure in a hard nonlinear algebraic system arising from realworld situations (Platt 2014). Kanagasabai and RavindhranathReddy (2014) proposed a combination of FPA and PSO for solving optimal reactive power dispatch problem. Sakib et al. 2014 compared FPA with Bat Algorithm. They tested these two algorithms on both unimodal and multimodal, low and high dimensional continuous functions and they observed that FPA gave better results.
3.2 Invasive Weed Optimization (IWO)
3.2.1 The inspiration phenomenon
IWO, inspired from the phenomenon of colonization of invasive weeds in nature, is proposed by Mehrabian and Lucas (2006). IWO is based on ecology and weed biology. It was seemed that mimicking invasive weeds properties, leads a powerful optimization algorithm. In a cropping field, weed colonization’s behavior can be explained as follows:
Weeds invade a cropping system by the way of disperse. They occupy suitable fields between plants. Each invasive weed takes the unused resources in the field, and becomes a flowering plant, and produces new invasive weeds independently. Number of new invasive weed produced by each flowering herb depends on fitness of flowering plants in the colony. These weeds provide better adaptation to the environment and grow faster by taking more unused resources and produce more seeds. The new produced weeds randomly spread over the field and grow to flowering plants. This process continues until reaching the maximum number of weed in the field because of limited resources. Weeds with better fitness can only survive and can produce new plants. The competition between weeds causes them to become well adapted and evolved over time (Karimkashi and Kishk 2010).
3.2.2 Algorithm
Some of the key terms used in IWO
Agent/seed  Each individual containing a value of each optimization variable in the colony 
Fitness  A value representing the goodness of solution for each seed 
Plant  A seed after evaluating its fitness 
Colony  all seeds or individuals 
Size of population  Number of plants in the colony 
Maximum number of plants  The maximum number of plants allowed to produce new seed in the colony 
 1.
Primarily, N parameters (variable) that need to be optimized must be selected. Then, in Ndimensional solution space, a maximum and a minimum value must be assigned for each of these variables (Defining the solution space) (Karimkashi and Kishk 2010).
 2.
A finite number of seeds are distributed randomly on the defined solution space. In another words, each seed randomly takes position in Ndimensional solution space. Position of each seed is an initial solution, which contains N values for the N variables, of optimization problem (Initializing a population).
 3.
Each of initial seed grows to a flowering plant. The fitness function, defined for representing goodness of the solution, returns a fitness value for each seed. Seed is called as a plant, after assigning the fitness value to the corresponding seed (Evaluate fitness of each individual) (Karimkashi and Kishk 2010).
 4.
Flowering plants are ranked according to fitness value assigned to them, before they produce new seeds. Then, each flowering plant produces seeds according to its ranking in the colony. In other words, the number of seed production of each plant depends on its fitness value or ranking and it increases from the minimum possible seeds (\(S_{min})\) to maximum possible seeds (\(S_{max})\). These seeds, which solve the problem better correspond the plants which are more adapted to the colony and therefore, they produce more seed. This step, by allowing all of the plants to participate in the reproduction contest, adds an important property to the algorithm (Ranking population and producing new seeds) (Karimkashi and Kishk 2010).
 5.In this step, produced seeds are spread to location of produced plant with equalaverage by normally distributed random numbers on the search space. At the present time step, the standard deviation (SD) can be expressed by (7).Here, \(iter_{max} \) is number of maximum iteration. \(\sigma _{iinitial} \) and \(\sigma _{final} \) are initial and final standard deviation respectively, and n is nonlinear modulation index. Algorithm starts with a high initial SD which can be explored by optimizer through the whole solution space. By increasing the number of iterations, for finding the global optimal solution, SD value is decreased gradually to search around the local minimum or maximum (Dispersion) (Karimkashi and Kishk 2010).$$\begin{aligned} \sigma _{iter} =\frac{\left( {iter_{max} iter} \right) ^{n}}{\left( {iter_{max} } \right) ^{n}}\left( {\sigma _{iinitial} \sigma _{final} } \right) +\sigma _{final} \end{aligned}$$(7)
 6.
New seeds grow to flowering plant, after all seeds have found their positions on the search space, and then, they are ranked with their parents. The plants in low rankings in the colony are eliminated for reaching number of maximum plant in the colony (\(P_{max} )\). The number of fitness evaluation, population size, is more than the maximum number of plants in the colony (Competitive exclusion) (Karimkashi and Kishk 2010).
 7.
Survived plants produce new seeds according to their ranking in the colony. The process is repeated at step 3, until the maximum number of iteration is reached or fitness criterion is met (Repeat) (Karimkashi and Kishk 2010).
3.2.3 Selection of control parameter values
Three parameters among all parameters affect the convergence of the algorithm; initial SS, \(\sigma _{iinitial} \), final SS, \(\sigma _{final} \), and nonlinear modulation index, n, must be tuned carefully for obtaining proper value of SS in each iteration according to (7). A high initial standard deviation must be chosen aggressively for allowing algorithm to discover the whole search space.
3.2.4 Studies with IWO
Mallahzadeh et al. (2008) used IWO for antenna configurations. Feasibility, efficiency and effectiveness of algorithm were examined with a set of antenna configurations for optimization of antenna problems. Karimkashi and Kishk (2010) applied IWO for different electromagnetic problems. They used IWO in linear array antenna synthesis, in the design of aperiodic thinned array antennas, and Uslot patch antenna for desired dualband characteristics. Rad and Lucas (2007) did work on recommender system which makes personalization according to features of userprofile. For each user, they used IWO to find the optimal priorities.
Roy et al. (2011) presented IWO for the design of nonuniform, planar, and circular antenna array which can achieve minimum side lobe levels for a specific first null beam width. For current application, they introduced more explorative routine which changes standard deviation of the seed population of classical IWO algorithm. Basak et al. (2010) used IWO to solve real parameter optimization problem which was related to design of timemodulated linear antenna arrays with Main Lobe Beam Width, ultralow Side Lobe Level, and Side Band Level. They included properties to classical IWO which were two parallel populations and a more explorative routine of changing the mutation stepsize with iterations. Monavar and Komjani (2011) proposed a new approach using Jerusalem crossshaped frequency selective surfaces as an artificial magnetic ground plane, for improving the bandwidth of a microstrip patch antenna. They used IWO to achieve optimal sizes of the JC_FSS element and the patch antenna.
Zaharis et al. (2012) introduced the improved adaptive beamforming technique of antenna arrays. This technique was implemented by IWO with adaptive dispersion. Unlike conventional IWO, seeds, produced by a weed, were distributed in the search space with a standard deviation, determined by fitness value of weed; and in this way they increased the convergence speed. Main purposes of Kostrzewa and Josiński (2009) study were to adapt the idea of the IWO to the problem of predetermining the progress of distributed data merging process, and to compare the results with the results obtained from evolutionary algorithm. Mehrabian and YousefiKoma (2007) developed a new approach for optimization of piezoelectric actuators in vibration suppression. They used IWO for maximizing of the frequency response function peaks, which reduced the vibration of smart fin. Ahmadi and Mojallali (2012) introduced a novel hybrid optimization algorithm using advantage of the stochastic properties of chaotic search and IWO. In order to deal with the weakness of conventional method, they presented chaotic IWO which incorporated the capabilities of chaotic search methods.
Nikoofard et al. (2012) presented a proposal for multiobjective IWO based on nondominant solutions ranking. This proposed algorithm was evaluated through a number of wellknown benchmarks for multiobjective optimization. Li et al. (2011) mentioned that there was no analytical formula for the YagiUda antenna design. Therefore, for getting the highest directivity, they integrated fullwave solver with IWO to optimize the variable parameters of YagiUda antenna. Pourjafari and Mojallali (2012) used IWO for solving nonlinear equations systems and they could find all real and complex root of a system. Ghasemi et al. (2014) presented a chaotic IWO algorithm based on chaos theory. They investigated its performance for control variables of optimal power flow with nonsmooth and nonconvex generator fuel cost curves.
3.3 Paddy Field Algorithm (PFA)
PFA was proposed by Premaratne et al. (2009) in 2009. PFA is a novel biology based algorithm which simulates growth process of paddy fields. Multiobjective optimization problem is considered as the growth process of rice in the PFA. Standard paddy field process can be divided into five parts: initialization, selection, sowing, pollination, and distribution. Considering a terminating condition, these five parts generate a cycle. When stopping criterion is reached, the cycle is terminated. Most seedproducing plant has a chance to survive and to reproduce. The selected plant offers the best solution in multiobjective optimization problems. Standard PFA was proposed to solve multiobjective optimization problem (Premaratne et al. 2009).
PFA works on principle of reproduction based on proximity to the general solution and population density similar to the plant population (Premaratne et al. 2009).
Assume that fitness function \(y=f\left( x \right) \), \(x=\left[ {x_1 ,x_2 ,\ldots ,x_n } \right] \). Each vector \(x=x_i \) is known as a seed or a plant, and its fitness value can be represented by y. Each dimensionality of seed must be in the range of \(x_j \in \left[ {a,b} \right] \).
Standard PFA, which simulates growth process, contains five main parts. These five main parts are: initiation, selection, sowing, pollination, and distribution. Bad seeds are eliminated at the end of an iteration count. After the good seeds are stored, better seeds are produced. Candidate solutions move toward the optimal solution during the cycle (Premaratne et al. 2009).
Step 1 Initialization
Each seed in the standard PFA represents a candidate solution. In a multiobjective optimization problem, seed is a vector. Algorithm initially works by distributing seeds randomly in the parameter space, to keep the seed variety (Premaratne et al. 2009).
Step 2 Selection
PFA is similar to GA. During evolution, populations, which cannot adapt to the environment, are eliminated. Generally, an objective function of a multiobjective problem is seen as a fitness function. After calculating fitness value of each seed, the best plants are selected according to a threshold value. nth most appropriate individual can be selected as the threshold operator. Plants will be eliminated whose fitness value is lower than the threshold value (Kong et al. 2012).
Step 3 Sowing
Step 4 Pollination
To obtain the number of neighbors of a plant, a sphere of radius a is used. Two plants such as \(x_j \) and \(x_k \) will be neighbors if they are in the sphere, and thus, neighbors’ number of each plant \(v_j \) can be determined. The maximum neighbors’ number of the plant in the population can be expressed as \(v_{max} \) (Kong et al. 2012).
3.3.1 Studies with PFA
Kong et al. (2012) reported that when the number of solutions was over the range in PFA, due to executing a lot of redundant iterations, the efficiency of algorithm became low. Therefore, they proposed a hybrid algorithm composed by PFA and Pattern Search Algorithm. Final results were found by pattern search based on the result of PFA. Wang et al. (2011) proposed PFA for the selection of Radial Basis Functions neural network center parameters. Cheng et al. (2011) used PFA for designing of PID controller of a highorder system.
3.4 Plant Root Growth
3.4.1 Root Mass Optimization (RMO) Algorithm
Root growth offers a great inspiration to propose a novel optimization algorithm. The objective function is considered as growth environment of the plant roots. The apices of the initial root forms a root mass. Each root apex can be considered as a solution of the problem. The roots turn the direction where provides optimal soil, water, and fertilizer conditions; thus they can reproduce. This process can be simulated as an optimizing process in the soil replaced with an objective function. According to this view, Qi et al. (2013) proposed RMO algorithm.
 (1)
All root apices forms a root mass. Two operators, including the root regrowth and the root branching, need to improve root growth behavior. Each root apex grows using one of these two operators.
 (2)Root mass are divided into three groups according to their fitnesses. The group with better fitness is called as regrowing group. The group with worse fitness, called as stop group, stops growing. The group, remaining from the root mass, is called as branching group. The meaning of the two operators, including root growth and root branching, are listed below (Qi et al. 2013).
 (a)Root Growth: This operator means root apex growth in the original direction. The root apex can migrate to the best position where provides the optimal soil water and fertilizer conditions. This operator can be formulated using (12).r is a random vector whose each element is in the range of [1, 1]. \(n_i \) is the new location of the ith root apex. \(x_i \) is the original position of the ith root apex. \(g_{best} \) is the root apex with the best fitness in each generation (Qi et al. 2013).$$\begin{aligned} n_i =x_i +r+\left( {g_{best} x_i } \right) \end{aligned}$$(12)
 (b)Root Branching: This operator means that root apex produces a new growth point instead of growing along the original direction. The growth point can be produced with a random angle \(\beta \), in a random position of original root. This operator can be formulated using (13).\(\alpha \) is a random number in the range of [0,1]. \(n_i \) is the new position of ith root apex. \(x_i \) is the original position of ith root apex. \(\beta \) can be calculated using (14).$$\begin{aligned} n_i =\beta \alpha x_i \end{aligned}$$(13)\(\lambda _i \) is a random vector (Qi et al. 2013).$$\begin{aligned} \beta =\lambda \Big /\sqrt{\lambda _i ^{T}\lambda _i } \end{aligned}$$(14)
 (a)
3.5 Artificial Plant Optimization Algorithm (APOA)
3.5.1 Main method
An important issue for simulating phenomena of plant growth is to peer this process into the optimization problem. Light intensity which guides to search direction in the problem space can be seen as fitness value due to guiding growth direction of plants and supplying significant energy by photosynthesis. Furthermore, a point can be seen as a branch and search strategy can be considered as growing curve. Because this new algorithm simulates growth model of plants which includes photosynthesis and phototropism mechanism, it is called briefly as APOA (Zhao et al. 2011).
3.5.2 Photosynthesis operator
Photosynthesis is a process which converts carbon dioxide to organic compounds (especially sugar) using the energy from sunlight. Photosynthesis occurs from algae in plants. Photosynthetic organisms are called as photoautotrophs as long as they established their own food. In plants and algae, photosynthesis uses carbon dioxide and water; it produces oxygen as a waste product.
Photosynthesis is vital for all aerobic life on Earth. In addition to maintaining the normal oxygen level in the atmosphere, almost all life is dependent to photosynthesis directly as an energy source or indirectly as an ultimate energy source of their food (chemoautotroph living in rock or around the deepsea hydrothermal vents) (Zhao et al. 2011).
3.5.3 Phototropism operator
Phototropism is usually seen in plants, but also can occur in other organisms like fungi. The cells on the plants far from the light, have a chemical that reacts when phototropism occurs. Therefore plants have elongated cells on the farthest side from the light. Phototropism is one of the tropism or movements that plants respond to external stimuli. Growth away from light is called as negative phototropism and growth towards a light source is a positive phototropism. Most plants exhibit positive phototropism, while their roots usually exhibit negative phototropism. Some vine shoot tips which allow growth towards to darkness, suspended solids and climbing, exhibit negative phototropism (Zhao et al. 2011).
3.5.4 Studies with APOA
The structure prediction of Toy model of protein folding is one of the bioinformatics problems. Cui et al. (2012) proposed APOA which could reach the local optimum easily and could find the global optimum with a larger probability. Furthermore, they used splitting strategy to improve the performance. Cui et al. (2013a) reported that in standard APOA, photosynthesis operator was selected only as rectangular hyperbolic, and some lightresponsive curves could increase the photosynthesis sensitivity. They used three different curves for it. In another study, they used Gravitropism mechanism (Cui et al. 2013b). Liu and Cui (2013) proposed a new hybrid APOA based Golden Section which increased the population density for protein folding problem. Furthermore; a famous local search strategy, Limited Memory Broyden–Fletcher–Shanno, was used to make an effective local search. Yu et al. 2013 proposed APOA with correlation branch which was no direct communication between branches unlike standard APOA.
Cai et al. (2012) chose seven classic models for research. These were rectangular hyperbolic model, nonrectangular hyperbolic model, updated rectangular hyperbolic model, parabolic model, straightline model and, two exponential curve models. Cai et al. (2013) designed binarycoded version of APOA and used this in HydrophobicPolar model which was simplest grid model in the protein structure prediction problem. Cui et al. (2014) designed a new dynamic local search strategy because exploitation capability is not suffice in APOA. Cai et al. (2014) indicated that in standard APOA there is no branching in each branch, on the contrary, in the real tree there are many branches in each branch. They incorporate two selection strategies into standard version for avoiding to this, and they tested it in DVHop algorithm. Zhao et al. (2011) proposed APOA to solve constrained optimization problems.
3.6 Sapling Growing up Algorithm (SGuA)
It is a computational method based on cultivating, growing up, and mating of saplings which is developed for searching and optimization problems. SGuA is discovered by Karci and his colleagues in 2006.
As a first step for this method, in sowing sapling, there must be equal length distance to each direction to each other. There are four operators for SGuA: Mating, Branching, Vaccinating, and Surviving (Karci 2007a, b, c; Karci and Alatas 2006).
The processes of SGuA can be explained by the following two phases:
Sowing Phase—Uniformed sampling: Saplings are scattered evenly to solution space.
Growing up Phase—This phase contains three operators: Mating operator aims to generate new saplings by mating of currently available saplings. Mating operator is a global search operator based on data exchange between two saplings (Karci 2007a, b, c; Karci and Alatas 2006).
Branching operator aims to generate new saplings from currently available saplings by using probabilistic method to determine the branching position of currently available saplings. Vaccinating operator aims to generate new saplings from currently available similar saplings. Vaccinating operator is a search operator which uses similar saplings (Karci 2007a, b, c; Karci and Alatas 2006).
3.6.1 Sowing saplings
In the first place, a solution space Z is defined as \(\hbox {Z}=\left\{ {z_0 ,\hbox {}z_1 ,\hbox {}\ldots ,\hbox {}z_{s1} } \right\} \) and it is assumed that this solution space is in binary. In this solution space, assuming that each sapling has two branches, then it is in the form \(\hbox {Z}=\left\{ {00,\hbox {}01,\hbox {}10,\hbox {}11} \right\} \) (Karci 2007a, b, c; Karci and Alatas 2006).
3.6.2 Growing up saplings
Growing up phase contains three operators: Mating, Branching, and Vaccinating. If the garden G(), contains G number of saplings, then the current garden is replaced by the G number of the best saplings of the offspring garden and the current garden. In other words, there isn’t any stochastic process in selection of the next generation garden (Karci 2007a, b, c; Karci and Alatas 2006).
(b) Branching: \(S_1 =\left\{ {s_{1,1} ,s_{1,2} ,\ldots ,s_{1,i} ,\ldots ,s_{1,n} } \right\} \) is a sapling. If one of the branch occurs in the point \(s_{1,i} \), then the probability of occurring of a branch in the point \(s_{1,i} \), \(i\ne j\), can be calculated in two ways: linear and nonlinear. The distance between \(s_{1,i} \) and \(s_{1,j} \) can be considered as \(\left {ij} \right \) or \(\left {ji} \right \). This process is shown in Fig. 12 (Karci 2007a, b, c; Karci and Alatas 2006).
3.6.3 SGuA
SGuA uses the similarity in the garden to generate the better saplings (Mating process). This is a cultural interaction process. Vaccinating process is the opposite of mating by the fact that vaccinating operator uses the dissimilarity in the garden. That is why; it is also the opposite of cultural interaction. Branching process is a unary operator which aims to embed the new information to the current solution set and to delete the information from the current solution set. Steps of SGuA are listed in the Fig. 14.
3.7 Photosynthetic Algorithm (PA)
Photosynthesis is one of the most important biochemical events. The most interesting photosynthetic reactions are considered as “dark reactions”. Dark reactions consist of a biochemical process which are combinations of Calvin–Benson cycle and photorespiration. The product of dark reaction is carbohydrates, like DHAP. PA uses the rules which regulate the conversion of carbon molecules from one substance to another in Calvin–Benson cycle, and photorespiration reactions. Replaced parts or shuffling are used to simulate the PA. PA is firstly proposed in 2000 by Murase. They chose analysis of the invert finite element to test the performance of PA as a typical optimization or parameter estimation (Murase 2000). Okayama and Murase (2002) chose Vizier problem as a typical optimization problem to show the performance of PA. Alatas (2011b) used PA in bioinformatics problems such as discovering association rules in biomedical data and multiplesequence alignment.
 (1)
Fixation of \(\hbox {CO}_{2}\) by ribulose bisphosphate carboxylase (Rubisco) and formation of 3 phosphoglycerate.
 (2)
Reduction of 3 phosphoglycerate to trioseP by the actions of glycerate3P kinase and NADPdependent glyceraldehydeP dehydrogenase.
 (3)
Regeneration of ribulose 1,5 bisphosphate by conversion of five \(\hbox {C}_{3}\) units into three \(\hbox {C}_{5}\) units (Alatas 2011b).
In the first stage of Calvin–Benson cycle, the rules that are regulating conversion of carbon molecules from one substance to another and the reaction located in chloroplast subcellular compartment for photorespiration are used by PA. DHAP, the product of photosynthesis, serves to provide information strings of the algorithm. When it is not possible to improve the quality of products, then the optimum point is reached. The quality of a product is evaluated according to the fitness value. Fig. 17 shows a flow diagram of the processes of the PA.
Algorithm starts with the random generation of intensity of light. Fixation rate of \(\hbox {CO}_{2}\) is evaluated according to the (32) which is based on the light intensity.
\(V_{max}\): Maximum \(\hbox {CO}_{2}\) fixation rate
A: Affinity of \(\hbox {CO}_{2}\)
L: Light intensity
All parameters that are given in (32) can be determined; however, these values can be assigned within a realistic range and they don’t need to be empirical. Light intensity might be randomly generated by the computer when PA is executed. Timevarying actual light intensity might be used as an alternative with online measuring system. For this phase, chaotic sequences may also be used. Change of light intensity as a stimuli is effective in reducing the occurring of local minimum traps in the search process. \(\hbox {CO}_{2}\) concentration in the leaf varies according to the \(\hbox {CO}_{2}\) fixation rate. \(\hbox {O}_{2}\) and \(\hbox {CO}_{2}\) concentration ratios are evaluated to determine the ratio of calculation frequency of Calvin–Benson cycle to the photorespiration cycle (Alatas 2011b).
3.8 Plant Growth Optimization (PGO)
PGO, which is simple but effective algorithm, is proposed to simulate plant growth with a realistic way considering the spatial occupancy, account branching, phototropism, and leaf growth. The main purpose of the model is to select active point by comparing the morphogen concentration for increasing the L system (Cai et al. 2008).
 1.
Produce new points with branching to search in optimal area which consists of the optimal solution;
 2.
The growth of leaves around the branching point to find accurate solutions in the local area;
3.9 Root Growth Algorithm (RGA)
3.9.1 Description for growth behaviors of root system
 (1)
A seed sprouts in the soil, and partly realizes growth as a plant stem above the soil surface. The other part of the seed grows downward to be the plant’s root system. The new root hairs grow from the root apex.
 (2)
The newer root hairs grow from the root apices of old root hairs. Repeated behavior of root system is called as branching of root apices.
 (3)
Most of the root hairs and root apices are similar. All of the plant’s root system has a selfsimilar structure. Root system of each plant consists of large number of root apices and root hairs with similar structure (Zhang et al. 2014).
3.9.2 Plant morphology
The effects of the root and rhizosphere characteristics on the plant resource efficiency are important. Uneven concentration of nutrients in the soil makes root hair growing in different directions. This characteristic of root growth is associated with the model of morphology in the biological theory. The formation of model can be considered as a complex process that cells are differentiated and create new spatial structure. When the rhizomes of the root system grows, three or four growth points with different rotation direction will be formed at each root apex. The rotation diversifies the growth direction of root apex. Root hairs contain cells whose root apices in the germinating place are undifferentiated. These cells are regarded as fluid bags in which there are homogeneous chemical components. One of the chemical components is a version of growth hormone called as morphactin. Morphactin concentration is a parameter of morphogenesis model. The parameter ranges between 0 and 1. The morphactin concentration determines whether cells start to divide or not. When cells start to divide, root hairs are visible (Zhang et al. 2014).
 (1)
If the plant root system has more than one root apex which can germinate, root hairs depends on their morphactin concentration. The probability of germination of new root hairs is higher from the root apices with larger morphactin concentration than root apices with less one.
 (2)
Morphactin concentration in the cell is not static, but it depends on its surrounding, in other words spatial distribution of nutrients in the soil. After the new root hairs germinate and grow, morphactin concentration will be reallocated between new root apices in parallel with new concentration of nutrients in the soil (Zhang et al. 2014).
3.9.3 Branching of the root tips
 (1)
Plant growth begins from seed.
 (2)
In each cycle of growth process, some excellent root apices which have larger morphactin concentration (fitness value in the embodied algorithm) are selected for branching.
 (3)
Distance should not be close between selected root apices in order to make spatial distribution wider and increase the diversity of fitness value.
 (4)
If number of selected root apices is equal to predefined value, the cycle of selection process ends (Zhang et al. 2014).
3.9.4 Root hair growth
 (1)If the morphactin concentration (fitness) of the new root apex is better than the old one in the same t cycle, the root apex will continue to grow. A new root apex in the inner loop can be expressed as:However, the number of iterations in the inner loop is a predefined value. If the number of iterations in the inner loop is equal to a predefined value, the inner loop stops (Zhang et al. 2014).$$\begin{aligned} x_i^t =x_i^t +\delta _i \varphi _i \end{aligned}$$(40)
 (2)If the morphactin concentration of the new root apex is worse than the old one in the same t cycle, the root apex will stop growing and \(t=\hbox {}t+1\). A new root apex can be expressed as:$$\begin{aligned} x_i^{t+1} =x_i^{t+1} +\delta _i \varphi _i \end{aligned}$$(41)
3.9.5 RGA
3.10 The Strawberry Algorithm as Plant Propagation Algorithm (PPA)
Strawberry plants belong to the rose family. Strawberrygrowing industry began in Paris in the 17th century by European diversity. AmedeFrancois Freizer, who was a mathematician and an engineer, was hired by Louis XIV in 1714 for drawing the map of South America, and he returned from Chile with Chilean strawberry plant which gives large fruits.
It is assumed that, strawberry plants, as well as other plants, have an underlying propagation strategy. This strategy is developed over time for ensuring to survive in species. In other words, a plant makes an effort to have children in areas which provide essential nutrients and growth potential. When a plant is in a good point, it sends lots of short runner, and when a plant is in a weak point, it sends longer runners for searching better points. Long runners are few, because they are investments. The plant may not have sufficient resource for sending many types of runners (Salhi and Fraga 2011).
It is predictable that with these observations and assumptions, strawberry plants and other plants can solve survival problem to grow in a particular environment. The mentioned approaches here contain mapping an optimization problem onto survival optimization problem of strawberry plant and adopting strategy of survival in the environment for searching points in the search space which give best values and ultimately the best value.
The strawberry algorithm is an exemplar of PPA. There are two features to emit all successful algorithms/heuristics for global optimization: concentration or intensification and diversification or exploration. While diversification provides searching to prevent jamming in a review of local optimum, concentration provides local search and convergence to a local optimum. This two features are contradictory and the success of any implementation of a search algorithm depends on the balance between these two features (Salhi and Fraga 2011).
This strawberry approach applies concentration by sending a large number of short runners from good solutions. It sends less but longer runner from the less good solutions to apply diversification. The full algorithm is presented in Fig. 21.
As many algorithms of this nature, PPA requires customization through the introduction of some auxiliary functions and assignment of values to a set of parameters. These parameters and functions for PPA are: size of population, a fitness function, the number of runners will be created for each solution, and distance for each runners (Salhi and Fraga 2011).
PPA is based on population of shoots each of which represents a solution in the search space. It is assumed that each shoot is taken to be equivalent to evaluated objective function. Each shoot will subsequently be sent to explore the solution space around it. The number of shoots is given by this parameter and it is shown as m in the algorithm. Algorithm is repeatedly run with sending runners from shoots in each generation. This parameter provides a termination condition based on how many times runners are transmitted, and is represented by \(g_{max} \).
3.10.1 Studies with PPA
Sulaiman et al. (2014) used modified PPA for constrained engineering optimization problems. Sulaiman et al. (2016) introduced a robust and efficient version of PPA and they provided more effective and improved variants of it on standard continuous optimization test problem in high sizes. Birds and animals provide dispersion of the seeds into the environment, while they are feeding from plants. Sulaiman and Salhi presented a seed based PPA by modelling this situation (Sulaiman and Salhi 2015).
3.11 Runner Root Algorithm (RRA)
RRA is inspired by plants such as strawberry and spider plants which are spread through their runners and also which develop roots and root hairs for local search for minerals and water resources. Similar to other metaheuristics, RRA does not apply same number of function evaluation at all iterations. More precisely, for optimal solution (exploitation procedure) in RRA, global search is performed at all iterations, while local search is performed only when global search does not lead to a significant improvement in the value of cost function. The only similarity between this algorithm and PPA is in using the idea of runners (MerrikhBayat 2015).

Each strawberry mother plant spreads through their runners. These runners rise randomly and each one leads to a new daughter plant (global search with big steps for resources).

Each strawberry mother plant improves its roots and root hairs randomly (local search with relatively big steps and small steps for resources).

Daughter strawberry plants which have access to rich resources, produce more runners and roots, grow fast, and, therefore, cover larger areas. On the other hand, daughter plants which move to poor resources will probably die (MerrikhBayat 2015).
3.11.1 Mathematical Description of RRA
Similar to other metaheuristic optimization algorithm, this algorithm, begins with a uniformly distributed random initial population, each called a mother plant in domain of the problem. The number of the mother plants is considered to be equal to \(N_{pop} \). Then, any mother plant at each iteration, except the fittest one, produces a daughter plant randomly in the domain of the problem (global search for better solutions). The distance of this each daughter plant from the mother plant is controlled by a parameter called \(d_{runner} \). The fittest mother plant produces a daughter plant exactly at the same location as itself. \(N_{pop} \) daughter plants are produced using this procedure. It is assumed that in RRA, at each iteration each mother plant is moved to a position referred by the daughter plant. If at least one of these daughter plants provides a significant improvement in the value of cost function compared with one of those obtained from the beginning of the algorithm, all of the mother plants required for the next iteration are selected from these daughter plants using a combination of elite and roulette wheel selection methods. More precisely, in this case, one of the necessary \(N_{pop} \) mother plants of the next iteration is considered to be equal to fittest daughter plant of the current iteration (elite selection), and the others are obtained by applying the roulette wheel method. In other words, while a significant decrease is observed in the value of cost function, the algorithm continuously examines the domain of the problem for better solutions through runners which improve only around the available solution. The amount of this improvement is measured by a parameter called tol.
However, if none of the resulting daughter plants leads to a pretty good value for cost function, algorithm starts to local search procedure through roots and root hairs. In this case, the algorithm assumes that the fittest daughter plant is in the same valley as the unknown global best solution. Therefore, if any improvements observed by applying a random change to each variable of this daughter plant, the algorithm accepts it. In other words, if the change which applies to some of the variables of the fittest daughter plant leads to a better solution, the algorithm accepts it and applies a random change to the other variable of the resulted daughter plant. If not, the previous value of this variable is retained and a random change is applied to the next variable of that daughter plant. This procedure is applied to all of the variables of the fittest daughter plant. This strategy guarantees that the general movement of the fittest daughter plant, which may be the global best solution, will be towards to the nearest minimum point in the same valley. From this case, the fittest daughter plant may be located in a valley that too flat or narrow, and both large and small random changes should be applied to each variable of the fittest daughter plant. In conclusion, in each iteration, a global search procedure is performed and a daughter plant is produced for all of the each mother plant. If none of the resulting daughter plant provides a significant improvement at the cost function value, the variables of the fittest daughter plant are subjected to random changes with both small and large steps (MerrikhBayat 2015).
The local search is not applied to all daughter plants which can significantly save the number of function evaluations and at all iterations. In addition to this, to increase the functionality, algorithm needs to be equipped with a restart strategy, because algorithm may be falling into the point trap of local optimum. For this purpose, after the global and potential local search, if the relative improvement is less than tol, value of the counter called stall_coune is incremented by 1; if not, value of the counter is equal to 0. If the value of this counter is equal to a predefined stall_max value, the algorithm is started with a random initial population. In other words, in this case, all the solutions obtained so far are discarded, and only the fittest solution obtained at the last iteration is memorized for comparison with the solutions previously or subsequently obtained to determine the best solution.
3.12 Path Planning Algorithm based on Plant Growth Mechanism (PGPP)
PGPP is established by extracting data processing mechanism in plant growth process. PGPP takes the plant bud as the main computing unit to search optimal plant. The growth of buds at different speeds and directions in PGPP embodies the group search capability. PGPP defines the seed germ as starting point and the light source as destination. PGPP discretizes plant growth period and replaces it with iterative calculations. During each iteration, it is assumed that plant growth behaviors remain the same. With the plant growth, the path which reaches the light source first is the best growth path. The path is planned according to this.
PGPP extracts phototropism, negative geotropism, apical dominance, and branching in plant growth as basic rules (Zhou et al. 2016).
Phototropism is the phenomenon consisting only towards the glare of one direction light source, and represents the plant adaptive mechanism to poor lighting. It is induced by the changes of the oxygen concentration in the light side and dark side. In PGPP, phototropism is summarized as follows: bud growth direction depends on the ambient light intensity in a limited reference area. Light intensity of calculation point is inversely proportional to the square of the distance between calculation point and destination point. Moreover, in order to pay attention to the shadow of the barrier in the reference range, the light intensity of the points under it requires a correction. The calculated light intensity is used for determining the growth rate and the growth direction of the bud.
Negative geotropism is the characteristic that front end of plant grows away from the ground and toward the sky. It is the complement of phototropism mechanism. Gravity vector introduces the effect of negative geotropism into PGPP. The line between the starting point and the ending point is defined as the vertical direction. Gravity vector becomes pointed from destination point to starting point. Both of negative geotropism and phototropism have an effect on the growth direction of the bud (Zhou et al. 2016).
Apical dominance is restraining effect on the growth in buds of lateral branches and is sprouting caused by apical bud of main branch which includes effect on the growth angle of lateral branches. It is due to the effect of two aspects. One of these is polarity of auxin transport. The other one is different response of the bud to auxin concentration. The polarity transport causes the buds of branches have a high oxygen concentration. Meanwhile, apical bud of main branch has appropriate auxin concentration to promote its growth, and apical dominance occurs. The transport amount of auxin polarity is determined by the differences between ages of branches.
3.12.1 Basic processes

Step 1: Initialization. Initialize variables such as plant and light intensity. Indicate the map data. Set the starting point as the growth point of the first bud.

Step 2: Calculation of light intensity. Calculate the light intensity in the reference range.

Step 3: Random branching. If the main branch reaches the branching condition, randomly create new branches.

Step 4: Calculation of growth vector. Calculate growth vector of light intensity and auxin concentration of the all buds. Then, to obtain the new growth vector, calculate weighted sum of the last period growth vector, gravitational growth vector, and light intensity growth vector.

Step 5: Plant growth. Each bud performs cell division until a new cell reaches the destination according to its growth vector. If any cell reaches the destination, route search ends and then proceeds to the next step. Otherwise, all the buds no sooner finish the growth of this period than go back to the second step.

Step 6: Way output. Way planning ends. The optimal way can be established from the start point to the destination point through being searched the parent cell by cell (Zhou et al. 2016).
3.13 Rooted Tree Optimization (RTO)
3.13.1 Roots look for water
A root has a limited capacity, but a group of roots can find together the best location to get water and their density positions around this location or around the way which connects the water resource with the plant. To create the algorithm, a presumptive behavior which is a way how the roots decide together, is added for selecting their orientation according to the wetness degree where the root head is in. These roots move randomly, but when one or more of them finds the wetness, to get the original location of the water and to strengthen their presence around this location, they call other roots. According to RTO algorithm, the solutions remote from water resources are omitted or replaced with the new roots oriented randomly. Moreover, the solutions remote from water resources can be changed by the roots which are next to the best root of the previous generation. The roots with significant wetness degree retain their orientation (Labbi et al. 2016).
3.13.2 RTO

Root: a candidate or proposed solution

Degree of wetness (\(D_w )\): a term that evaluates a candidate and gives fitness degree between population segments (Labbi et al. 2016).
3.13.2.2 The rate of the continuous root in its orientation (\(R_c \))
3.13.2.3 The rate of the random root \((R_r \))
\(R_n ,R_r ,\) and \(R_c \) ratios are experimentally determined according to the investigated problem. These ratios are considered as variables affecting the convergence. \(R_r \) value is always small compared with others, because it aims to reserve the randomness to be far from local solutions. \(D_w \) value is added from the search function of roots for determining a search area according to candidate force (Labbi et al. 2016).
3.13.3 RTO algorithm

Step 1: Randomly create the first generation which consists of N candidates within variable limits in the search area, and set the numerical values of \(R_n \), \(R_r \), and \(R_c \) ratios.
 Step 2: Evaluate the whole population members for measuring \(D_w \) as following:or directly use the fitness regardless of the suitable formula.$$\begin{aligned} D_w \left( \hbox {i} \right) =\left\{ {{\begin{array}{l} \frac{\hbox {f}_i }{\hbox {max}\left( {\hbox {f}_i } \right) }\quad \hbox {for the maximum objective} \\ 1\frac{\hbox {f}_i }{\hbox {max}\left( {\hbox {f}_i } \right) }\quad \hbox {for the minimum objective} \\ \end{array} }} \right. ,\quad i=1,2,\ldots ,N \end{aligned}$$(57)

Step 3: Reproduce and replace with the new population. According to \(R_n \), \(R_r \), and \(R_c \) reorder the population using the (54)(56), according to wetness degree to replace with the new population.

Step 4: If the termination conditions are not reached, go back to step 2 (Labbi et al. 2016).
4 Discussions
Depending on the source of inspiration, all of the plant based metaheuristic algorithms in the literature have been searched and collected for the first time to present a relatively comprehensive list and to inspire further research. One of the research works in the related literature consist of comparing one of these plant based algorithms to the other nature inspired algorithms. These works have experimentally shown that some plant intelligence based algorithms are better than other nature inspired methods although it is still not quite understood why. Especially, the older plant intelligence based algorithms have been compared with GA and PSO. Algorithms which lack certain basic capabilities such as the mixing and diversity among the solutions can be thought as badly designed. Most of the plant based metaheuristic search and optimization algorithms which have been searched and collected for the first time in this paper have been developed relatively more recently and there is not any work about comparing all of their performances in the selected complex benchmark or engineering problems satisfying the equal conditions. Equal conditions mean using the same starting and termination criterion, equal number of starting search points, the same benchmark function with equal dimensions and intervals for these dimensions, same hardware running the algorithms, same programming languages, and maybe the same programmer. Thus, it can be concluded that there is not a single software available for evaluating all these new plant based algorithms over different types of benchmark functions; this provides an opportunity for the research community to develop the unified software that could be used to simulate the performance of these algorithms over benchmark problems used for search and optimization.
Software which can be developed in near future can also be extended for solving various other types of discrete, combinatorial, NP hard, multiobjective, complex optimization problems etc. It can be concluded that metaheuristic algorithms inspired from plant behaviors have much room to grow since this research community is quite young. Thus, this paper would act as a boon to the research community in identifying the research prospects in the field of metaheuristic optimization.
If more works are performed in relatively more recent plant based algorithms, they can have both mixing and diversity control so that the algorithms can explore the vast search space efficiently, while converge relatively more quickly. Some good and relatively old algorithms such as FPA and SGuA have both global search and intensive local search capabilities, which balance the exploration and exploitation, and they seem partly more efficient in unimodal and multimodal functions or problems. Their solution quality in terms of mean objective value is much better and the value of the standard deviation indicates that the results obtained from them are more consistent due to the huge number of works on their performances. More works should be performed for more recent algorithms such as RTO and PGPP which have been proposed in 2016, RRA and PPA which have been proposed in 2015 in order to increase the performance in terms of convergence, precision, robustness, and general performance.
In future, more comprehensive evaluation of all plant inspired metaheuristic algorithms with graphical and tabular analysis may be focused on. Their variants containing multi objectivity, dynamic parameter selection, different initial population methods, different termination criterion, discrete versions, and their hybrids with other metaheuristic or heuristic methods may be proposed for efficient solutions as these plant based metaheuristic algorithms have high potential to solve various search and optimization problems in different areas. Researchers should be encouraged to implement a detailed performance analysis of these algorithms truly to pick up the best methods for different types of hard problems. Therefore, searching and collecting all of the plant based metaheuristic algorithms in the literature for the first time and inspiring more research to gain better insight into efficient algorithms and solve largescale realworld problems has been aimed with this paper.
5 Conclusions
Although there are many successful search and optimization algorithms and techniques in the literature; design, development, and implementation of new techniques is an important task under the philosophy of improvement in the scientific field and always searching to design better. Best algorithm that gives the best results for all of the problems have not yet been designed, that is why constantly new artificial intelligence optimization algorithms are proposed or some efficient additions or modifications have been performed to the existing algorithms.
In recent years, researchers have shown that plants exhibit intelligent behaviors. Although there are thirteen different plant intelligence based metaheuristic search and optimization algorithms, they are not known by the researchers of the related area. However, they seem to be very popular in solving of highdimensional and hard problems due to their efficiency and robustness.
In this paper, all of the metaheuristic algorithms inspired from plant intelligence have been firstly searched, collected, and their properties are introduced in a smooth way. In this way, the main algorithm and source of inspiration are introduced. Besides, the basic steps of the plant intelligence based optimization algorithms are demonstrated. Consequently, studies about related algorithms are shown to reveal how those algorithms helped solving of complex problems.
This paper will help the related researchers to rapidly gain information about novel techniques, the differences between these techniques, and adopting them for different types of complex search and optimization problems.
References
 AbdelRaouf O, AbdelBaset M, Elhenawy I (2014a) An improved flower pollination algorithm with chaos. Int J Edu Manage Eng 4:1CrossRefGoogle Scholar
 AbdelRaouf O, AbdelBaset M, Elhenawy I (2014b) A new hybrid flower pollination algorithm for solving constrained global optimization problems. Int J Appl Op Res 4:1–13CrossRefGoogle Scholar
 AbdelRaouf O, ElHenawy I, AbdelBaset M (2014c) A novel hybrid flower pollination algorithm with chaotic harmony search for solving sudoku puzzles. Int J Mod Edu Comput Sci 6:38CrossRefGoogle Scholar
 Ahmadi M, Mojallali H (2012) Chaotic invasive weed optimization algorithm with application to parameter estimation of chaotic systems. Chaos Solitons fractals 45:1108–1120MathSciNetCrossRefGoogle Scholar
 Alatas B (2011a) ACROA: artificial chemical reaction optimization algorithm for global optimization. Expert Syst Appl 38:13170–13180CrossRefGoogle Scholar
 Alatas B (2011b) Photosynthetic algorithm approaches for bioinformatics. Expert Syst Appl 38:10541–10546CrossRefGoogle Scholar
 AtashpazGargari E, Lucas C (2007a) Imperialist competitive algorithm: an algorithm for optimization inspired by imperialistic competition. In: 2007 IEEE congress on evolutionary computation. 25–28 Sept. pp 4661–4667. doi: 10.1109/CEC.2007.4425083
 AtashpazGargari E, Lucas C (2007b) Imperialist competitive algorithm: an algorithm for optimization inspired by imperialistic competition. In: IEEE congress on evolutionary computation. CEC 2007, pp 4661–4667. doi: 10.1109/CEC.2007.4425083
 Basak A, Pal S, Das S, Abraham A, Snasel V (2010) A modified invasive weed optimization algorithm for timemodulated linear antenna array synthesis. In: IEEE congress on evolutionary computation (CEC). 18–23 July. pp 1–8. doi: 10.1109/CEC.2010.5586276
 Borji A (2007) A new global optimization algorithm inspired by parliamentary political competitions. In: MICAI 2007: advances in artificial intelligence. Springer, pp 61–71. doi: 10.1007/9783540766315_7
 Cai W, Yang W, Chen X (2008) A global optimization algorithm based on plant growth theory: plant growth optimization. In: 2008 International conference on intelligent computation technology and automation (ICICTA). 20–22 Oct. pp 1194–1199. doi: 10.1109/ICICTA.2008.416
 Cai X, Fan S, Tan Y (2012) Light responsive curve selection for photosynthesis operator of APOA. Int J BioInspired Comput 4:373–379. doi: 10.1504/IJBIC.2012.051411 CrossRefGoogle Scholar
 Cai X, Li P, Wu X (2014) Artificial plant optimization algorithm with double selection strategies for DVHop. Sensor Lett 12:1383–1387. doi: 10.1166/sl.2014.3356 CrossRefGoogle Scholar
 Cai X, Wu X, Wang L, Kang Q, Wu Q (2013) Hydrophobicpolar model structure prediction with binarycoded artificial plant optimization algorithm. J Comput Theor Nanosci 10:1550–1554. doi: 10.1166/jctn.2013.3439 CrossRefGoogle Scholar
 Cheng M, Dawei D, Pandeng Z, Jia L, Fei L (2011) A novel method using PFA in parameters turning of PID controller of highorder system. In: 2011 IEEE international conference on information and automation (ICIA). 6–8 June, pp 662–665. doi: 10.1109/ICINFA.2011.5949076
 Chu SC, Tsai PW, Pan JS (2006) Cat swarm optimization. In: PRICAI 2006: trends in artificial intelligence. Springer. pp 854–858. doi: 10.1007/9783540366683_94
 Cui Z, Fan S, Zeng J, Shi Z (2013) Artificial plant optimisation algorithm with threeperiod photosynthesis. Int J BioInspired Comput 5:133–139. doi: 10.1504/IJBIC.2013.053507 CrossRefGoogle Scholar
 Cui Z, Liu D, Zeng J, Shi Z (2012) Using splitting artificial plant optimization algorithm to solve toy model of protein folding. J Comput Theor Nanosci 9:2255–2259. doi: 10.1166/jctn.2012.2647 CrossRefGoogle Scholar
 Cui Z, Liu W, Dai C, Chen W (2014) Artificial plant optimization algorithm with dynamic local search for optimal coverage configuration. Sensor Lett 12:118–122. doi: 10.1166/sl.2014.3238 CrossRefGoogle Scholar
 Cui Z, Liu X, Liu D, Zeng J, Shi Z (2013) Using gravitropism artificial plant optimization algorithm to solve toy model of protein folding. J Comput Theor Nanosci 10:1540–1544. doi: 10.1166/jctn.2013.3437 CrossRefGoogle Scholar
 Cura T (2008) Modern sezgisel teknikler ve uygulamaları. Papatya Yayıncılık,Google Scholar
 Dorigo M, Maniezzo V, Colorni A (1991) The ant system: an autocatalytic optimizing process. Technical ReportGoogle Scholar
 Geem ZW, Kim JH, Loganathan G (2001) A new heuristic optimization algorithm: harmony search. Simulation 76:60–68CrossRefGoogle Scholar
 Ghasemi M, Ghavidel S, Akbari E, Vahed AA (2014) Solving nonlinear, nonsmooth and nonconvex optimal power flow problems using chaotic invasive weed optimization algorithms based on chaos. Energy 73:340–353. doi: 10.1016/j.energy.2014.06.026 CrossRefGoogle Scholar
 Ghasemi M, Ghavidel S, Rahmani S, Roosta A, Falah H (2014b) Novel hybrid algorithm of imperialist competitive algorithm and teaching learning algorithm for optimal power flow problem with nonsmooth cost functions. Eng Appl Artif Intell 29:54–69. doi: 10.1016/j.engappai.2013.11.003 CrossRefGoogle Scholar
 Ghasemi M, Taghizadeh M, Ghavidel S, Abbasian A (2016) Colonial competitive differential evolution: an experimental study for optimal economic load dispatch. Appl Soft Comput 40:342–363. doi: 10.1016/j.asoc.2015.11.033 CrossRefGoogle Scholar
 Ghasemi M, Taghizadeh M, Ghavidel S, Aghaei J, Abbasian A (2015) Solving optimal reactive power dispatch problem using a novel teachinglearningbased optimization algorithm. Eng Appl Artif Intel 39:100–108. doi: 10.1016/j.engappai.2014.12.001 CrossRefGoogle Scholar
 He S, Wu QH, Saunders JR (2006) A novel group search optimizer inspired by animal behavioural ecology. In: 2006 IEEE international conference on evolutionary computation. 1272–1278. doi: 10.1109/CEC.2006.1688455
 He S, Wu QH, Saunders JR (2009) Group search optimizer: an optimization algorithm inspired by animal searching behavior. IEEE Trans Evolut Comput 13:973–990. doi: 10.1109/TEVC.2009.2011992 CrossRefGoogle Scholar
 Holland J (1975) Adaption in natural and artificial systems. The University of Michigan Press, Ann ArborMATHGoogle Scholar
 Jin X, Reynolds RG (1999) Using knowledgebased evolutionary computation to solve nonlinear constraint optimization problems: a cultural algorithm approach. In: 1999 IEEE congress on evolutionary computation, CEC 99. vol. 3. pp. 1678. doi: 10.1109/CEC.1999.785475
 Kanagasabai L, RavindhranathReddy B (2014) Reduction of real power loss by using fusion of flower pollination algorithm with particle swarm optimization. J Inst Ind Appl Eng 2:97–103Google Scholar
 Karaboğa D (2011) Yapay Zeka Optimizasyon Algoritmaları. Nobel Yayın DağıtımGoogle Scholar
 Karci A (2007a) Human being properties of saplings growing up algorithm. In: IEEE international conference on computational cybernetics. ICCC 2007. 19–21 Oct. pp 227–232. doi: 10.1109/ICCCYB.2007.4402039
 Karci A (2007b) Natural inspired computational intelligence method: saplings growing up algorithm. In: IEEE international conference on computational cybernetics. ICCC 2007. 19–21 Oct. pp 221–226. doi: 10.1109/ICCCYB.2007.4402038
 Karci A (2007c) Theory of saplings growing up algorithm. In: adaptive and natural computing algorithms. lecture notes in computer science. Berlin: Springer. Heidelberg. pp 450–460. doi: 10.1007/9783540716181_50
 Karci A, Alatas B (2006) Thinking Capability of saplings growing up algorithm. In: Intelligent data engineering and automated learning—IDEAL 2006. vol 4224. Lecture notes in computer Science. Springer, Berlin, pp 386–393. doi: 10.1007/11875581_47
 Karimkashi S, Kishk AA (2010) Invasive weed optimization and its features in electromagnetics. IEEE Trans Antennas Propag 58:1269–1278. doi: 10.1109/TAP.2010.2041163 CrossRefGoogle Scholar
 Kashan AH (2009) League championship algorithm: a new algorithm for numerical function optimization. In: IEEE international conference of soft computing and pattern recognition. SOCPAR’09. pp 43–48. doi: 10.1109/SoCPaR.2009.21
 Kaveh A, Talatahari S (2010) A novel heuristic optimization method: charged system search. Acta Mechanica 213:267–289. doi: 10.1007/s0070700902704 CrossRefMATHGoogle Scholar
 Kennedy J, Eberhart R (1995) Particle swarm optimization. In: IEEE international conference on neural networks. Piscataway, NJ. pp 1942–1948Google Scholar
 Kong X, Chen YL, Xie W, Wu X (2012) A novel paddy field algorithm based on pattern search method. In: 2012 International conference on information and automation (ICIA). 6–8 June. pp 686–690. doi: 10.1109/ICInfA.2012.6246764
 Kostrzewa D, Josiński H (2009) The comparison of an adapted evolutionary algorithm with the invasive weed optimization algorithm based on the problem of predetermining the progress of distributed data merging process. In: ManMachine Interactions. vol 59. Adv Intel Soft Comput. Berlin: Springer Heidelberg. pp 505–514. doi: 10.1007/9783642005633_53
 Labbi Y, Attous DB, Gabbar HA, Mahdad B, Zidan A (2016) A new rooted tree optimization algorithm for economic dispatch with valvepoint effect. Int J Electr Power Energy Syst 79:298–311. doi: 10.1016/j.ijepes.2016.01.028 CrossRefGoogle Scholar
 Lam AYS, Li VOK (2010) Chemicalreactioninspired metaheuristic for optimization. IEEE Trans Evolut Comput 14:381–399. doi: 10.1109/TEVC.2009.2033580 CrossRefGoogle Scholar
 Lenin K (2014) Shrinkage of active power loss by hybridization of flower pollination algorithm with chaotic harmony search algorithm. Control Theory Inform 4:31–38Google Scholar
 Li Y, Yang F, OuYang J, Zhou H (2011) YagiUda antenna optimization based on invasive weed optimization method. Electromagnetics 31:571–577. doi: 10.1080/02726343.2011.621108 CrossRefGoogle Scholar
 Liu D, Cui Z (2013) Protein folding structure prediction with artificial plant optimization algorithm based golden section and limited memory Broyden–Fletcher–Goldfarb–Shanno. J Bionanosci 7:114–120. doi: 10.1166/jbns.2013.1103 CrossRefGoogle Scholar
 Łukasik S, Kowalski P (2015) Study of flower pollination algorithm for continuous optimization. In: Intelligent systems’2014. vol 322. Advances in intelligent systems and computing. Springer International Publishing. pp 451–459. doi: 10.1007/9783319113135_40
 Mallahzadeh ARR, Oraizi H, DavoodiRad Z (2008) Application of the invasive weed optimization technique for antenna configurations. Prog Electromag Res 79:137–150. doi: 10.2528/PIER07092503 CrossRefGoogle Scholar
 Maniezzo V, Stützle T, Voss S (2009) Matheuristics: hybridizing metaheuristics and mathematical programming, vol 10. Springer, New YorkMATHGoogle Scholar
 Mehrabian AR, Lucas C (2006) A novel numerical optimization algorithm inspired from weed colonization. Ecol Inform 1:355–366. doi: 10.1016/j.ecoinf.2006.07.003 CrossRefGoogle Scholar
 Mehrabian AR, YousefiKoma A (2007) Optimal positioning of piezoelectric actuators on a smart fin using bioinspired algorithms. Aerosp Sci Technol 11:174–182. doi: 10.1016/j.ast.2007.01.001 CrossRefGoogle Scholar
 MerrikhBayat F (2015) The runnerroot algorithm: a metaheuristic for solving unimodal and multimodal optimization problems inspired by runners and roots of plants in nature. Appl Soft Comput 33:292–303. doi: 10.1016/j.asoc.2015.04.048 CrossRefGoogle Scholar
 Monavar FM, Komjani N (2011) Bandwidth enhancement of microstrip patch antenna using jerusalem crossshaped frequency selective surfaces by invasive weed optimization approach. Prog Electromagn Res 121:103–120. doi: 10.2528/PIER11051305 CrossRefGoogle Scholar
 Murase H (2000) Finite element inverse analysis using a photosynthetic algorithm. Comput Electr Agr 29:115–123. doi: 10.1016/S01681699(00)001393 CrossRefGoogle Scholar
 Nikoofard AH, Hajimirsadeghi H, RahimiKian A, Lucas C (2012) Multiobjective invasive weed optimization: application to analysis of pareto improvement models in electricity markets. Appl Soft Comput 12:100–112. doi: 10.1016/j.asoc.2011.09.005 CrossRefGoogle Scholar
 Okayama T, Murase H (2002) Solution for NQueens problem using a photosynthetic algorithm. In: 15th Triennial World Congress IFACGoogle Scholar
 Platt GM (2014) Computational experiments with flower pollination algorithm in the calculation of double retrograde dew points. Int Rev Chem Eng 6(2):95–99Google Scholar
 Pourjafari E, Mojallali H (2012) Solving nonlinear equations systems with a new approach based on invasive weed optimization algorithm and clustering. Swarm Evolut Comput 4:33–43. doi: 10.1016/j.swevo.2011.12.001 CrossRefGoogle Scholar
 Prathiba R, Moses MB, Sakthivel S (2014) Flower pollination algorithm applied for different economic load dispatch problems. Int J Eng Technol 6:1009–1016Google Scholar
 Premaratne U, Samarabandu J, Sidhu T (2009) A new biologically inspired optimization algorithm. in: 2009 international conference on industrial and information systems (ICIIS). 28–31 Dec. pp 279–284. doi: 10.1109/ICIINFS.2009.5429852
 Qi X, Zhu Y, Chen H, Zhang D, Niu B (2013) An idea based on plant root growth for numerical optimization. In: Intelligent computing theories and technology. vol 7996. Lecture notes in computer science. Springer, pp 571–578. doi: 10.1007/9783642394829_66
 Rad HS, Lucas C (2007) A recommender system based on invasive weed optimization algorithm. In: IEEE congress on evolutionary computation, CEC 2007. 25–28 Sept. pp 4297–4304. doi: 10.1109/CEC.2007.4425032
 Rao RV, Savsani VJ, Vakharia DP (2011) Teachinglearningbased Optimization: a novel method for constrained mechanical design optimization problems. ComputerAided Des 43:303–315. doi: 10.1016/j.cad.2010.12.015 CrossRefGoogle Scholar
 Rashedi E, NezamabadiPour H, Saryazdi S (2009) GSA: a gravitational search algorithm. Inf Sci 179:2232–2248. doi: 10.1016/j.ins.2009.03.004 CrossRefMATHGoogle Scholar
 Roy GG, Das S, Chakraborty P, Suganthan PN (2011) Design of nonuniform circular antenna arrays using a modified invasive weed optimization algorithm. IEEE Trans Antennas Propag 59:110–118. doi: 10.1109/TAP.2010.2090477 CrossRefGoogle Scholar
 Sakib N, Kabir M, Subbir M, Alam S (2014) A comparative study of flower pollination algorithm and bat algorithm on continuous optimization problems. Int J Appl Inf Syst 7:19–20Google Scholar
 Salem SA (2012) BOA: a novel optimization algorithm. In: IEEE 2012 International Conference on engineering and technology (ICET). pp 1–5. doi: 10.1109/ICEngTechnol.2012.6396156
 Salhi A, Fraga ES (2011) Natureinspired optimisation approaches and the new plant propagation algorithm. In: The international conference on numerical analysis and optimization (ICeMATH ’11). Yogyakarta, IndonesiaGoogle Scholar
 ShahHosseini H (2009) The intelligent water drops algorithm: a natureinspired swarmbased optimization algorithm. Int J BioInspired Comput 1:71–79. doi: 10.1504/IJBIC.2009.022775 CrossRefGoogle Scholar
 Simon D (2008) Biogeogr Based Optim. IEEE Trans Evolut Comput 12:702–713. doi: 10.1109/TEVC.2008.919004 CrossRefGoogle Scholar
 Storn R, Price K (1995) Differential evolution: a simple and efficient adaptive scheme for global optimization overcontinuous spaces. Technical Report TR95012. vol 3. ICSI BerkeleyGoogle Scholar
 Sulaiman M, Salhi A (2015) A seedbased plant propagation algorithm: the feeding station model. Sci World J 2015:16. doi: 10.1155/2015/904364 Google Scholar
 Sulaiman M, Salhi A, Fraga ES, Mashwani WK, Rashidi MM (2016) The plant propagation algorithm: modifications and implementation. Sci Int 28Google Scholar
 Sulaiman M, Salhi A, Selamoglu BI, Kirikchi OB (2014) A plant propagation algorithm for constrained engineering optimisation problems. Math Probl Eng 2014:10. doi: 10.1155/2014/627416 CrossRefGoogle Scholar
 Sundareswaran K, Nayak PS, Sankar P, Kumar VV (2014) Inverter harmonic elimination through flower pollination enhanced genetic algorithm. Int J Adv Trends Comput Sci Eng 3:342–348Google Scholar
 Wang GG, Deb S, Cui Z (2015) Monarch butterfly optimization. Neural Comput Appl. doi: 10.1007/s005210151923y
 Wang R, Zhou Y (2014) Flower pollination algorithm with dimension by dimension improvement. Math Probl Eng. doi: 10.1155/2014/481791
 Wang S, Dai D, Hu H, Chen YL, Wu X (2011) RBF neural network parameters optimization based on paddy field algorithm. In: 2011 IEEE international conference on information and automation (ICIA). 6–8 June. pp 349–353. doi: 10.1109/ICINFA.2011.5949015
 Yang XS (2012) Flower Pollination Algorithm for global optimization. In: Unconventional computation and natural computation. Springer. pp 240–249. doi: 10.1007/9783642328947_27
 Yang XS, Karamanoglu M, He X (2013) Multiobjective flower algorithm for optimization. Procedia Comput Sci 18:861–868. doi: 10.1016/j.procs.2013.05.251 CrossRefGoogle Scholar
 Yang XS, Karamanoglu M, He X (2014) Flower pollination algorithm: a novel approach for multiobjective optimization. Eng Optim 46:1222–1237. doi: 10.1080/0305215X.2013.832237 MathSciNetCrossRefGoogle Scholar
 Yang XS, Suash D (2009) Cuckoo search via levy flights. In: World Congress on nature & biologically inspired computing. NaBIC 2009. 9–11 Dec. pp 210–214. doi: 10.1109/NABIC.2009.5393690
 Yu B, Cui Z, Zhang G (2013) Artificial plant optimization algorithm with correlation branches. J Bioinform Intel Control 2:146–155. doi: 10.1166/jbic.2013.1039 CrossRefGoogle Scholar
 Zaharis ZD, Skeberis C, Xenos TD (2012) Improved antenna array adaptive beamforming with low side lobe level using a novel adaptive invasive weed optimization method. Prog Electromag Res 124:137–150. doi: 10.2528/PIER11120202 CrossRefGoogle Scholar
 Zhang H, Zhu Y, Chen H (2014) Root growth model: a novel approach to numerical function optimization and simulation of plant root system. Soft Comput 18:521–537. doi: 10.1007/s005000131073z CrossRefGoogle Scholar
 Zhao Z, Cui Z, Zeng J, Yue X (2011) Artificial plant optimization algorithm for constrained optimization problems. In: 2011 Second international conference on innovations in bioinspired computing and applications (IBICA). 16–18 Dec.. pp 120–123. doi: 10.1109/IBICA.2011.34
 Zhou Y, Wang Y, Chen X, Zhang L, Wu K (2016) A Novel path planning algorithm based on plant growth mechanism. Soft Comput 1–11: doi: 10.1007/s005000162045x