Evolutionary Parameter Tuning Method for a Graph-Based Path Planning Algorithm in Agricultural Robotic Manipulators

The successful execution of path planning algorithms can be substantially impacted by the selection of appropriate planner’s parameters. Identifying the optimal parameters suitable for a specific task and environment may require a heuristic process. For an agricultural robotic manipulator operating in a dynamic environment, achieving the optimal balance to satisfy many performance criteria can be challenging. This work addresses the issue of parameter tuning for a graph-based path planning algorithm called the Bi-directional Transition-based Rapidly Exploring Random Tree (BiTRRT) algorithm in a given high-dimensional search space environment. The objective of this work is to propose a highly efficient evolutionary tuning method, based on a genetic algorithm, that optimizes the parameters of the BiTRRT algorithm, taking into account the features and characteristics of a given task and working environment. The performance of the robotic manipulator is measured by the generated paths that minimize both planning and execution time. A genetic algorithm and a high-fidelity simulation are used together in the proposed method. This combines the benefits of a stochastic optimization approach with a substantial degree of real-world replication, facilitating the fine-tuning of the optimization process. The genetic algorithm was designed with a distinctive fitness function that encompasses the total path cycle time accumulated from various motion scenarios. A high-resolution three-dimensional occupancy grid map of a tree was created from stereo-depth camera pointcloud data for the high-fidelity simulation. Experimental results indicated that the BiTRRT algorithm, when optimized with optimal parameters, outperforms the baseline BiTRRT algorithm, yielding a 5.7% - 7.6% enhancement in the completion of path planning tasks varying in geometric complexity. Given the improvement in a robotic manipulator’s agricultural service life, operational productivity may substantially increase.