One of the fundamental challenges of the twenty-first century is to create a food- and nutrition-secure society, as the world’s population approaches 10 billion by 2050 and 11 billion by 2100 [1], without exhausting Earth’s biological and natural resources [24]. It is no wonder that all key aspects of the food systems (FS) are under constant pressure to become more sustainable, less polluting, efficient, and resilient and to move towards zero food waste, while providing affordable food which is healthy, safe, of consistent quality, and even personalized to individual consumers [47]. All aspects of the food system must be adapted, from food production to food consumption, which includes the entire food supply chain. This article focuses mostly on the latter.

Demands on the FS present challenges and opportunities to the supply chain actors, including the efficient use and smart optimization of current processes (e.g., to reduce food and ingredient waste, to use energy efficiently, to minimize production costs), offset labor shortages (e.g., Brexit-related shortages of international workers), provide human-contact-free food handling (e.g., to reduce the risk of pathogen transmission), and to provide personalized food (i.e., customized for target groups and individuals). Automation and robotic technologies are expected to play a major role in addressing these issues [711, 12••, 13•, 14], which is evident in the growing number of robotic installations across the food industry [15, 16].

This article examines robotic advances and their potential to address post-harvest supply chain challenges for fresh produce such as fruits and vegetables (Fig. 1). Fresh produce, by its nature, is perishable. The primary role of the post-harvest sector is to ensure fresh product quality from farm to fork. Still, a significant portion of the food is lost or wasted in all steps of the chain, reaching over 30% in high, middle, and lower income countries, almost half of which is lost prior to retail [1719]. To minimize these losses, it is paramount to have suitable harvesting protocols, grading-sorting processes, and organization of the transport chain and storage conditions [17,18,19,20]. In the post-harvest supply chain, fresh product quality management plays a prominent role. From the chain performance perspective, an integrative view of logistics and product quality has been considered essential by several researchers [21,22,23,24]. Robotics and automation are expected to offer a multitude of opportunities in food supply chain logistics, such as increased productivity, higher efficiency, resource optimization, end-to-end tracking and traceability, improved warehouse logistics, and reduced health and safety hazards [9, 11, 13•, 25, 26, 27•].

Fig. 1
figure 1

Representation of the roles of robotics in the post-harvest supply chain. The post-harvest supply chain consists of Quality control & sorting shortly after harvest, Storage on location or during transportation, Pick & pack to prepare produce for the next actor in the chain, Distribution to retail, and Retail itself. Robotic technologies can support multiple activities in the chain, including manipulation operations (such as grading, sorting, packaging, palletizing) and storage management and material flow (using mobile robots for, i.a. real-time monitoring of product quality and storage conditions, automated inventory and stock management, efficient space organization)

Monitoring quality and safety, making decisions, and taking actions along the different parts of the post-harvest chain require different approaches that depend on the individual chain elements. Robotics can play a vital role in addressing such challenges in the post-harvest chain. The role of advances in robotics is reported here with a focus on monitoring, maintaining, and preserving product quality, as well as facilitating quality-driven decision-making in the post-harvest supply chain.

In the following sections, the current state and the outlook of robotics for post-harvest quality management are detailed with emphasis on two key perspectives: manipulation and navigation. Both perspectives emphasize varying degrees of interaction between the robot and its environment, ranging from the grasping and manipulation of objects to navigation and transport. Subsequent sections attend to robotic technologies suitable for grasping, manipulation, and movement of fresh products across the post-harvest supply chain. The conclusions are reported in the closing section. To keep the survey size manageable, contact-free sensing technologies and artificial intelligence, though highly relevant, are not included in the scope of the survey.

Robotic Manipulation

One of the most common set of robotic activities in post-harvest processes is repetitive pick-and-place operations, which range from grading/sorting actions to processing, packaging, and palletizing [9, 12••]. A key distinction to be considered here is when manipulation requires direct contact with the fresh product (grading, sorting, cutting, primary packaging) against indirect contact (e.g., stacking and palletizing).

Indirect Contact

Indirect contact primarily involves grasping, moving, and placing uniform and rigid objects, albeit of different weight, shape, and size, such as individual packages or crates stacked on pallets. Such applications are prevalent from the moment of harvest (stacking crates with harvest), through transport and storage, all the way to retail. These are some of the more common applications with several exiting industrial robotic solutions [9, 12••] (Fig. 2). Such robots are generally not specific to food industry, but their deployment needs to consider which robots fit the existing factory floor planning, payload capacity, and right end-effectors to handle the diversity (in size, shape, weight, intact/deformable) of packages to be handled. Therefore, the current need of the industry is to have a structured approach to industrial robot selection suited to their current processes. In a recent work, Bader and Rahimifard propose the 4-step FIRM (Food Industrial Robot Methodology) approach for industrial robot selection in food processing operations [12••]. The scope of their work is broader than post-harvest products and processes, but the methodology is highly relevant for processes where industrial robots already exist.

Fig. 2
figure 2

Examples of industrial robot solutions for object handling where the robot does not directly make contact with the product inside the packages. a ABB robotic palletizer IRB 660 (source: ABB Robotics); b FANUC M410-iB mixed palletizing solution (source: FANUC); and, c Depalletizing robot solution from SERFRUIT (source: SERFRUIT)

Direct Contact

In the post-harvest supply chain, direct contact between the robot and the products occurs primarily during quality control (e.g., grading and sorting) and pick-place operations (e.g., primary packaging). The main point of contact is between the produce and the robotic end-effector (a.k.a. end-of-arm tooling), which can grab, hold, and release objects, as well as move the objects to a desired location. Although a variety of robotic handling solutions have been proposed in industry as well as academia, robotic handling approaches have been slow to catch on in the food industry. Contributing factors include societal issues such as economic and social barriers, as well as technological limitations for flexible grasping and manipulation, and adverse effects of direct contact on product organoleptic quality and overall hygiene [9, 12••, 28, 29]. Here, we focus on the technological perspective.

Robotic grasping and manipulation actions that require direct contact with the fresh products pose two primary challenges:

  • Products are difficult to grasp and manipulate: complex and varied characteristics of the products (irregular shapes and sizes (e.g., ranging from rigid cabbage to soft and fragile berries) and high deformability (e.g., a bunch of grapes)) make them extremely difficult to grasp and manipulate. Several works have extensively categorized foodstuff characteristics and the challenges they pose for manipulation [12••, 30•, 31•].

  • Grasping and manipulation may damage the product. Improper handling exposes fresh produce to damage and pathogen spread, which affect the overall post-harvest quality, leading to changes not only in appearance, but also in other organoleptic properties such as taste, smell, and texture [8, 12••, 30•].

The traditional, and most prominent, approaches for robotic handling of fresh products in the industry are contact-based (e.g., electric, pneumatic, hydraulic, inflating rubber, multi-body mechanism grippers) and air-based (e.g., suction cups commonly coupled with contact-based grippers, Bernoulli principle-based levitation grippers) [12••, 28, 29, 31•]. Other mechanisms, such as pinching and freeze gripping that are common for foodstuff manipulation, are generally not suitable for fruits and vegetables since the gripping principles are inherently destructive to fresh products. Adding compliant materials to gripping elements (e.g., rubber pads) is a widely employed, but an incomplete solution to the problem.

Given the difficulty to handle a variety of fresh products and the potential to damage by grasping, traditional end-effectors are suitable only for a limited number of products. That is, the solutions are not universal. In particular, handling of fragile products is extremely challenging with traditional end-effectors. With advances in materials science and soft components, there is an increasing trend towards lighter, simpler, and more generalized grasping solutions [31•, 32, 33]. In the design of such universal grippers, advanced materials are increasingly studied that can change shape in response to external or internal actuators as well as to their interactions with objects [32, 34]. Application of these materials makes grippers capable of grasping and manipulating a wide array of objects of different shapes, sizes, and rigidity. Additionally, increased mechanical compliance also means that the control complexity for these grippers is highly reduced. In the last decade, several such solutions, primarily silicone elastomer-based grippers, have transitioned from research labs to commercial products [35]. Due to strict hygiene requirements, not all of these end-effectors are suitable for fresh products (or foodstuff, in general). In recent years, several food-grade soft end-effectors have been commercialized which are waterproof and cleanable for foodstuff handling (Fig. 3).

Fig. 3
figure 3

Examples of soft end-effectors capable of handling irregular and fragile products, and meeting hygiene standards for post-harvest applications. a FinRay gripper “DHAS” from Festo modeled on the fin of a fish (source: Festo); b OnRobot Softgripper, a flexible food-grade gripper which does not require external air supply for grasping and releasing actions (source: OnRobot)

Soft robotics grasping and manipulation remains a highly active field of research with inspiration coming from i.a. material properties (such as silicone elastomers, shape memory materials, active polymers, and gels) [32], biology [33, 36], and folding and cutting patterns [37, 38]. These approaches will play a key role in handling soft and delicate products and are highly relevant for a multitude of fresh products not manipulable via traditional rigid-joint end-effectors.

In terms of robotic handling, significant progress has been reported on grasping and manipulation mechanisms in academia as well as industry. For post-harvest problems, the key utility of these mechanisms lies in their ability to not deteriorate the fresh product quality, neither at individual level (e.g., due to bruising, dropping, leaving residues), nor at the batch level (carrying over contaminants from one product to another). In the last decade, several researchers have proposed mappings of gripping technologies to foodstuff handling [12••, 28, 29]. These mappings offer broad guidelines by grouping large families of gripping methods and their potential pros and cons for foodstuff handling. However, there is a lack of precise benchmarking, for instance to quantify the impact of specific end-effectors on a product’s storage life, retail shelf life, pathogen spread, and organoleptic properties. As the number and variety of robotic grasping techniques continues to increase, benchmarking procedures are essential to determine the suitability of grasping and manipulation materials and techniques according to post-harvest quality criteria.

Robotic Navigation

Another family of robotic solutions that offers new opportunities at various points in the post-harvest chain are the autonomous mobile (navigable) robots, AMRs, that can be applied to optimize transport and material flows [11, 13•, 25, 27•]. In particular, AMRs can play a significant role in facilitating and optimizing autonomous quality monitoring and quality-centric decision-making.

Quality Sensitive Warehouse Automation

Perishable goods have short lives and storage conditions have a great impact on product quality and eventual shelf life. For instance, product-specific controlled atmosphere storage can preserve shelf life and quality for longer when compared to retail warehouses. For perishable products, inventory and warehouse management need to explicitly consider product perishability and deterioration, quality-driven demand forecasting, pricing and stock management, food loss prevention, and increasingly the sustainability demands [2439,40,41,42, 43••]. These constraints result in a significant amount of human movement and material flow in fresh produce warehouses, specifically at the tasks of monitoring, visiting storage shelves, picking goods, and replenishing shelves. In time-critical operations, inefficiency of the pickers’ movements has been considered a major disadvantage in traditional warehouses [44]. Consequently, there has been a focus on developing approaches to warehouse automation, including the use of AMRs, for autonomous storage and retrieval, as well as for storage (re)organization, particularly in e-retail and manufacturing [25, 44•, 45, 46••, 47,48,49,50]. The use of AMRs is highly relevant and applicable to fresh produce warehousing.

Autonomous Mobile Robots

AMRs are autonomous and unmanned mobile robots capable of reacting to their surroundings, follow the required trajectory based on facility maps, dynamically respond to and navigate around obstacles, move at desired speeds, and haul varying payloads. Automated-guided vehicles (AGVs) are a subset of AMRs which can also navigate through the facilities and transport payloads; however, AGVs can only navigate through fixed trajectories and react to obstacles by halting (no other evasive action is possible) [45]. For the purposes of this article, we take an inclusive definition of AMRs, but it should be noted that, industrially, AMRs and AGVs are considered different automation solutions.

The most common application of an AMR is assisted picking. In assisted picking, the robot either follows the picker to a picking location or travel to the picking location and wait for pickers to arrive, after which the pickers load the robot with the desired goods [44•, 45, 47, 49, 51]. The AMRs then navigate back to a central depot where the goods are unloaded (Fig. 4a). A Robotic Mobile Fulfillment System (RMFS) consists of a swarm of mobile robots, each capable of lifting movable racks in warehouses and deliver them to workers [44•, 45, 47, 48, 49, 52]. Unlike static racks in traditional warehouses, these movable and portable racks represent a very different approach to warehouse management. A key feature of RMFS is its ability to sort inventory automatically and to adapt the warehouse layout quickly. AVGs also play an important role in automated storage and retrieval systems, such as Robot-based Compact Storage and Retrieval (RCSR). In a RCSR system, goods-holding bins are stored in a highly dense manner with a grid overlaid on top. Mobile robots roam on top of the grid and have the capability to lift and extract bins and transport them to workstations [46••, 47, 48, 53] (Fig. 4b, c).

Fig. 4
figure 4

Examples of different types of AMRs in warehousing applications. a MiR industrial AMR for internal transportation (source: MiR100, Honeywell Safety & Productivity Solutions, UK); b CarryPick RMFS from Swisslog (source: Swisslog); and c AutoStore RCSR system (source: AutoStore)

RMFS and RCSR systems are a complete rethink of warehousing systems, with robotic automation being the central objective. AMR-assisted picking, on the other hand, can be retrofitted to existing warehouses. Besides robotic solutions, there are several other approaches being explored for warehouse automation, for which readers are directed to the comprehensive surveys of Boysen et al. 2019 [44•] and Azadeh et al. 2019 [47].

Despite the potential, these technologies are currently not widely applied in the logistics sector (including food logistics) due to concerns relating to rapidly evolving technologies, associated costs, changing actors in the omnichannel supply chain, the risks associated with short-term contracts (prevalent in multi-actor supply chains), and lack of understanding of the technology [11, 13•, 27, 44•].

A Quality-Centric View

Existing industrial AMRs already offer a wide range of standard applications in supporting warehouse automation, although their use continues to be underexplored in the fresh produce industry. Adoption of these technologies, in their current form, can already benefit post-harvest warehousing solutions. Such advantages can range from efficient storage and retrieval solutions, efficient material flows, automated inventory management, less hazardous working conditions, and handling labor shortages. This section explores AMRs beyond these standard applications, especially from the fresh produce quality perspective.

It is critical to understand why quality monitoring is essential during storage. Harvested fresh produce is still alive and thus respires and produces heat. Increased temperatures increase respiration rates and thus enhance the rates of deterioration and ripening processes. Additionally, the autocatalytic fruit ripening hormone ethylene, which i.a. is induced by stress, also leads to enhanced respiration [54, 55]. Unchecked, large quantities of fresh produce can be wasted in a matter of days or a few weeks. Stress, for instance, leads to ripening, increased respiration, and increased temperature, which in turn causes the surrounding produce to ripen, potentially causing a chain reaction. To minimize respiration, several strategies are employed during storage, such as reducing temperatures, lowering oxygen (O2), increasing carbon dioxide (CO2), and avoiding stress [56,57,58]. Added to this, different fresh products have different storage requirements, which do not always coincide with the conditions present at the warehouses or may conflict with fluctuations in supply and demand. Real-time monitoring of the changing climate conditions as well as product quality is extremely challenging in fresh produce warehouses. Moreover, while low temperatures, low O2 and high CO2 conditions are good for prolonged storage of fresh produce, they also carry considerable risks for human employees that are engaged in regular quality monitoring tasks. Currently, to perform these tasks, conditions are relieved temporarily or quality monitoring frequency is reduced.

AMRs, due to their high payload capacity, ability to autonomously traverse warehouses and transport goods (typically packaged products, bins, or racks), provide new opportunities for monitoring fresh produce quality and making quality-driven decisions. With high payload capacity, new sensing equipment can be mounted on AMRs, autonomous navigation allows the robots to easily navigate the warehouse, while actively measuring the environment with climate sensors. Contact with products enables sensors to measure quality traits and climate conditions in real-time. Their transportation capability allows AMRs to move products autonomously based on product quality, environmental conditions, and required changes due to fluctuations in warehouse stock.

The availability of quality and climate data in real-time and the ability to physically act upon that information can transform inventory management and warehouse operations. The advantages can range from quality-based dynamic stock and warehouse management, quality-driven demand forecasting, dynamic sales decisions (e.g., soon –to-be-perishable product selected for a closer market), dynamic pricing based on current quality, early detection and removal of deteriorating, damaged, or diseased products, and maintenance of uniform storage conditions specific to the product requirements. Additionally, the data can be used to facilitate and optimize mixed storage, i.e., storage of fresh produce with different storage requirements, by rearranging inventory along temperature and atmospheric gradients in the storage space.

Combining quality aspects with robotic system concerns such as identifying optimal operational policies (e.g., robot (re)routing, collision avoidance, robot zoning, order picking, monitoring), design optimization (e.g., storage layout, number of robots, number and location of workstations, storage rack arrangement), and overall system performance (e.g., in terms of throughput, sustainability, health, and safety) presents a new set of research and operational opportunities.

A Comment on Transport

A critical part of the post-harvest supply chain is transportation of goods. During transport, transportation units are moved around and are subject to impact damage, depending on the packing of the produce and the infrastructure present during the transport. During this period, quality monitoring is limited mainly by the tight packing of produce to maximize use of the storage space. In some cases, this is further restricted by the low temperature and the controlled atmosphere conditions in the transportation units, typically shipping containers. This setting is not conducive to current AMR-based solutions due to the comparatively large size of the robotic units and the highly compact storage arrangement. A more suitable solution would be Internet of Things (IoT)-enabled sensors, mounted inside transport containers, which continuously monitor both climate and product quality [59,60,61,62].


This article reviews the current state of robotic manipulation and navigation technologies and their potential for quality management and quality-driven decision-making in post-harvest supply chains.

Robotic Manipulation: Current State and Future Perspectives

Robotic handling of fresh produce remains an open problem, especially for irregular, deformable, and fragile products. Advances in soft robotics are leading to the next generation of universal gripping solutions which are soft, mechanically compliant, and with low control complexity. However, for fresh produce, successful grasping is a partial solution. Product handling should also not deteriorate the product (e.g., due to bruising, dropping, leaving residues, transferring contaminant). A key future research direction is to consider an integrative approach to grasping and quality preservation. A lack of precise benchmarking procedures is also evident from the literature. With the proliferation of robotic grasping techniques, there is an urgent need for benchmarking procedures according to post-harvest quality criteria.

Robotic Navigation: Current State and Future Perspectives

Automated mobile robots, or AMRs, are increasingly being used in manufacturing and e-commerce warehousing applications. In post-harvest warehouse operations, their use is less common. Adoption of AMR technologies, in their current form, can already benefit post-harvest warehousing solutions by optimizing material flow, automating storage and retrieval, assist in picking and autonomous inventory management among other advantages. A quality centric view is proposed, where beyond their standard use, AMRs carry additional sensors to monitor warehouse climate and fresh produce quality in real-time and execute quality-driven (semi) autonomous actions (e.g., re-arrange stock, re-route bin for sale, extract affected bins, flag affected area). Future research should focus on leveraging AMRs for quality management and decision-making, in addition to supporting warehouse automation.