Resilient Timed Elastic Band Planner for Collision-Free Navigation in Unknown Environments
Geesara Kulathunga,Abdurrahman Yilmaz,Zhuoling Huang,Ibrahim Hroob,Hariharan Arunachalam,Leonardo Guevara,Alexandr Klimchik,Grzegorz Cielniak,Marc Hanheide
2024-12-04
Abstract:In autonomous navigation, trajectory replanning, refinement, and control command generation are essential for effective motion planning. This paper presents a resilient approach to trajectory replanning addressing scenarios where the initial planner's solution becomes infeasible. The proposed method incorporates a hybrid A* algorithm to generate feasible trajectories when the primary planner fails and applies a soft constraints-based smoothing technique to refine these trajectories, ensuring continuity, obstacle avoidance, and kinematic feasibility. Obstacle constraints are modelled using a dynamic Voronoi map to improve navigation through narrow passages. This approach enhances the consistency of trajectory planning, speeds up convergence, and meets real-time computational requirements. In environments with around 30\% or higher obstacle density, the ratio of free space before and after placing new obstacles, the Resilient Timed Elastic Band (RTEB) planner achieves approximately 20\% reduction in traverse distance, traverse time, and control effort compared to the Timed Elastic Band (TEB) planner and Nonlinear Model Predictive Control (NMPC) planner. These improvements demonstrate the RTEB planner's potential for application in field robotics, particularly in agricultural and industrial environments, where navigating unstructured terrain is crucial for ensuring efficiency and operational resilience.
Robotics,Optimization and Control