rA*: Re-Planned A* Technique for Point-to-Point Robot Navigation in Dynamic Environments
Tanudeep Ganguly, Rapti Chaudhuri, Suman Deb · 2024
Autonomous point-to-point robots employ path-finding algorithms to traverse from the start to the goal point by avoiding obstacles. One such prominent and efficient algorithm is A*. It uses the best first technique to traverse a particular environment and extract the shortest path by avoiding on-path obstacles. It is widely accepted for its optimality in finding the shortest path from one node to another. In the case of real-world scenarios, robot navigation involves dynamic obstacles with varying costs and multiple objectives. This paper proposes the rA* (replanned A*) technique, which implements important modifications to the A* algorithm to address the challenges created in dynamic environments by introducing dynamic cost adjustment with sensor data to reduce path distance in replanning in case of obstacles or complex terrain features differences in an environment. We also enhance the heuristic approach with environmental awareness by equipping Euclidean distance for faster navigation. Real-time and multi-objective path planning have been conceptualized to dynamically adjust path in case of moving obstacles. Real-time analysis of the algorithm is conducted on a Turtlebot by taking the sensor-fused input data in coordinate and depth format. The sensor data is obtained through ROS (Robot Operating System) visualizer and the map of the path exploration by detecting the dynamic obstacles is generated through G-mapping. The maps are fed as input to the rA* algorithm to re-plan the path, distorted by the sudden appearance of dynamic obstacles. rA* reduces the overall and near-obstacles turning angle of the robot by 80.4% - 82.2%, as compared to the traditional A* technique. The results prove rA* to be superseded in terms of accurate path tracking and control by precise obstacle avoidance and achieving intelligent navigation.