A Convex Optimization Based Differentially Driven Mobile Robot Planner for Crowd Navigation
Leixin Chang, Haoran Yuan, Tengyue Wang, Haonan Mai, L. Yang · 2024
Safe navigation in a pedestrian-rich environment has gained a lot of attention in robotics research. Unlike classical motion planning with a static environment, pedestrian-rich scenarios are associated with a highly dynamic environment and safety risk. In this paper, we propose a novel local planner designed for differentially driven wheeled robots, directly inspired by the Dynamic Window Approach (DWA). Our model leverages convex optimization and differential drive kinematics to efficiently determine optimal velocity inputs as the robot moves in the human crowd. Our approach does not only account for the position but also the velocity of the pedestrians in the planning framework to facilitate safer navigation through dynamic pedestrian-dense environments. Through extensive simulation experiments, we demonstrate the superior effectiveness and safety of our method compared to DWA, showcasing significant enhancements in collision-free navigation success rates and computational efficiency through the use of convex optimization techniques. Code release: https://github.com/Leixinjonaschang/convex_op_planner.