Performance Analysis of Path Planning Algorithms for Fruit Harvesting Robot

PurposePath planning is an essential part in designing of an agricultural robot. The path planning algorithms for fruit harvesting robots vary in performance, based on different environments, obstacles, and constraints. This research aims to analyze and evaluate the most commonly used path planning algorithms by fruit harvesting robots in the past 10 years to assess the robot's performance. The primary objective behind the comparative analysis of path planning algorithms is to ascertain which algorithm demonstrates better performance in terms of reaching the target fruit in the shortest time, requiring the least amount of computing resources, and being able to navigate around obstacles effectively. Hence, the study determines which path planning algorithm is the most efficient for the application of fruit harvesting robot.MethodIn this study, four common path planning algorithms were evaluated namely A-star, Probabilistic Road Map, Rapidly exploring Random Tree, and improved Rapidly exploring Random Tree. Three cases were examined for performance. The first case deals with performance based on varying orientations of fruit within the workspace. The second case investigates the performance in the presence of obstacles in the path, and the third case caters to performance due to varying distances of robot and the fruit. Matlab software was used for creating simulation environment for testing. Run time, path length, standard deviation, and total task time were obtained for each case and statistical analysis was done.ResultsIt was found that improved Rapidly exploring Random Tree performed better in terms of path length and gave an optimal path as compared to the other algorithms due to its rewiring feature by an average of 21%. Run time of Rapidly exploring Random Tree was better than the other three algorithms.ConclusionFour most commonly used path planning algorithm were analyzed for performance for fruit harvesting robot for three different cases. Despite the variations in performance across different scenarios, the results confirmed that the improved Rapidly exploring Random Tree algorithm outperformed all other algorithms under the given constraints.