AD-Based Dynamically Feasible Replanning Technique for Anguilliform-Inspired Robot

• Published in: Journal of intelligent & robotic systems Vol. 108; no. 3; p. 53
• Main Authors: Ojha, Pritam, Thakur, Atul
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.07.2023; Springer Nature B.V
• Subjects: Artificial Intelligence; Barriers; Collision avoidance; Collision dynamics; Control; Electrical Engineering; Engineering; Fish; Fluid flow; Mechanical Engineering; Mechatronics; Path planning; Regular Paper; Robot dynamics; Robotics; Robots; Success; Dynamic obstacles; Fish-inspired robots; Anytime dynamic A; Replanning; Path planning; Anguilliform-inspired robot
• ISSN: 0921-0296; 1573-0409
• DOI: 10.1007/s10846-023-01902-6

This paper presents an AD*-based technique for fast, dynamically feasible path replanning for collision avoidance of anguilliform-inspired articulated mobile robots operating in the presence of both static and dynamic obstacles in underwater environments. Although a few past approaches have addressed the problem of the generation of dynamically feasible paths in the presence of static obstacles, to the best of our knowledge, path planning in the case of fish-inspired robots in the presence of dynamic obstacles has not been explored. We employ a lattice-based approach to generate dynamically feasible reachable nodes and, thereby, a search graph that captures the robot’s kinodynamic constraints and accounts for the interaction with surrounding fluid flow. We employ proximity sensors to sense the dynamic obstacles for replanning. Moreover, a simple vector-based time-to-collision and collision state prediction method has been proposed to adapt the problem to a graph-search approach. We found that the developed approach could avoid collision with dynamic obstacles moving at 1.5 times the robot’s speed, with a maximum success rate of around 78%. The average success rate obtained using the developed technique is 70%, which is more than the success rate obtained using D*-Lite for the examples reported in this paper. The developed approach can impart autonomy to fish-inspired robots operating in unstructured and cluttered environments.