Multi-agent Task Assignment Using Swap-Based Particle Swarm Optimization for Surveillance and Disaster Management

Abstract

In this paper, a multi-agent task assignment for simultaneous tasks is proposed which is suitable for surveillance and disaster management. The service requests which are represented as the tasks from the areas of interest include respective GPS coordinates from the Google Maps. The multi-agent system is assigned with these tasks following a two-stage approach. At first, the tasks are distributed for each agent based on the proximity of an agent to tasks, inter-task proximity, and task completion overhead. The assigned tasks of each agent are then orchestrated as the traveling salesman problem. A swap-based particle swarm optimization (PSO) is proposed for optimizing the sequence of executions of the assigned tasks of each agent. The proposed method is demonstrated on areas whose selections are inspired by real-life disasters such as the Tsunami-affected Marina Beach at Chennai, India and the earthquake-affected M. G. Marg at Gangtok, India. Results show the suitability of the proposed method for such multi-task multi-agent system.

5 Aerial Distance Between Two GPS Coordinates

The proposed method is implemented on Google Maps. To calculate the aerial distance between two GPS coordinates, the law of cosines model is used, where it is assumed that earth is spherical [27] and this model considers all the GPS coordinates at the mean sea level. The following process is used for the calculation of aerial distance. Let ${\textit{GPS}}_1^{o}~=~({\textit{lat}}_1^{o},{\textit{lon}}_1^{o})$, where ${\textit{lat}}_1^{o}$ and ${\textit{lon}}_1^{o}$ are in decimal degree. Then,

$$\begin{aligned} {\textit{GPS}}_1^{r}&={\textit{GPS}}_1^{o}~\times ~\left( \displaystyle \frac{\pi }{180}\right) \nonumber \\ {\textit{GPS}}_2^{r}&={\textit{GPS}}_2^{o}~\times ~\left( \displaystyle \frac{\pi }{180}\right) \nonumber \end{aligned}$$

Let $A~=~{\textit{sin}}({\textit{lat}}_1^{r})\times ~{\textit{sin}}({\textit{lat}}_2^{r})$ and $B~=~{\textit{cos}}({\textit{lat}}_1^{r})\times ~{\textit{cos}}({\textit{lat}}_2^{r})\times ~{\textit{cos}}({\textit{lon}}_2^{r}~-~{\textit{lon}}_1^{r})$, then $\text {aerial~distance}=$ ${\textit{cos}}^{-1}(A+B)\times ~\text {earth~radius}$. To obtain accuracy in few meters, ${\textit{cos}}^{-1}$ needs to be accurate up to 10 decimal places or in double format.