On the Performance Analyses of a Modified Force Field Algorithm and Neural Network Approach for Obstacle Avoidance in Swarm Robotics

Abstract

Obstacle avoidance is a major hurdle when implementing mobile robots and swarm robots. Swarm robots work in groups and therefore require an efficient and functional obstacle avoidance algorithm to stay collision free between themselves and their surroundings. This paper reviews previous research in obstacle avoidance implementation using the force field method (FFM), also known as potential field method (PFM) and a neutral network approach. Moreover, this paper aims to execute simulations using a modified force field algorithm and a neural network approach and compare them. The obtained results are analyzed to identify the performance characteristics and the time taken to perform tasks using a singular mobile robot against a swarm robot environment consisting of four and ten robots, respectively, in both simulation cases. Simulations showed that the algorithm was successful in navigating obstacles for both single and swarm robot environments. A single robot was found to take up to 340% longer to arrive at the required location compared to the first robot in the experiment. Moreover, it was found that the neural network approach showed ~ 27% improvement over the modified force field algorithm when it comes to cases where more than four robots are being used.