OMEPP: Online Multi-Population Evolutionary Path Planning for Mobile Manipulators in Dynamic Environments

Yangjun Pi, Xin Liu, Zuodong Yang, Yunlin Zhong, Tao Huang, Huayan Pu, Jun Luo

Abstract: This paper presents an online multi-population evolution path planning (OMEPP) algorithm to address the flexible path planning problem for mobile manipulators in complex dynamic environments. The OMEPP algorithm treats the mobile manipulator as a high-dimensional system to utilize its flexibility. The OMEPP algorithm is based on random sampling and evolutionary concepts: Optimization and passive obstacle avoidance operations are performed on the path at runtime, with superior paths replacing inferior ones within the same population. A novel path population partitioning approach is proposed to maintain diverse switchable paths, thereby improving robustness. This paper also proposes an efficient manipulator collision detection method and several other mechanisms to enhance the algorithm’s effectiveness. The experimental results demonstrate the algorithm’s ability to swiftly adapt and optimize paths in response to dynamic environmental changes. Note to Practitioners—This paper presents OMEPP, an online evolutionary algorithm for real-time path planning of mobile manipulators in dynamic environments. OMEPP employs novel techniques including path population partitioning, random sampling, and evolution to efficiently generate collision-free paths among moving obstacles. A novel path population partitioning approach is proposed to maintain diverse switchable paths, thereby improving robustness. Simulations have demonstrated that the OMEPP algorithm is effective for real-time path planning of mobile manipulators in complex dynamic environments. Future work will focus on trajectory generation respecting dynamics limits.