An Enhanced Timed Elastic Band Method for Autonomous Navigation and its Collision Avoidance Reliability Analysis

Abstract

Abstract The timed elastic band (TEB) method is an optimization-based navigation algorithm which uses the idea of an elastic band to connect an agent’s current pose to its goal position. The method is composed of three main levels: i) a feasible global path, ii) deforming collision-free path in real time, and iii) a control method to move the robot along the path. Static and moving obstacles deform the elastic band but the agent will always be pulled toward the goal position while avoiding obstacles. The key issue is the required computation efficiency to calculate the collision-free path in real time under dynamically changing environment. In literature, various case studies have been implemented to demonstrate the performance considering various constraints. However, its collision avoidance reliability under safety critical navigation scenarios has not yet been studied. The major contributions of this paper are summarized as follows. Firstly, an enhanced version of the TEB (i.e., eTEB) is proposed to improve the computational efficiency by explicitly considering non-holonomic constraints of the TEB method. Secondly, collision avoidance reliability analysis of the eTEB method is systematically conducted considering non-reactive moving obstacles. Results indicate that the eTEB method is much more reliable than the DWA for head-on collision avoidance with a single non-reactive moving obstacle at a time.