Planning-based optimized path for automatic robot navigation

Abstract

Automatic navigation is one of the primary features of a fully autonomous robotic system i.e, achieving the goal without human intervention. Laser-based navigation requires map of the environment in order to move around. The map not only imposes memory overhead but also asserts processing load for its access all the time. Moreover Laser sensor is an expensive device to be used. Instead we present a low-cost, simple, and faster approach for robot navigation based on PDDL-based planning. A wheeled mobile humanoid robot having one functioning arm, built for navigation and communication with other differential drive robots. Robot uses 2m/s as translation velocity and π/8 rad/s as rotational velocity. Ultrasonic sensor has been used for obstacle avoidance. In order to automatically navigate a PDDL-based path planner is used that requires a problem with initial and goal states and a set of actions as input. Planner generates a path plan e.g., {turn-left, turn-left, move-forward} which is also shared with other robots. The robot environment is represented as state space where robot's pose (position+orientation) is state, and connection between states are the actions Left, Right, and Forward. Our PDDL-based plan execution time is 30% time less than that of Logic-based planner time. Robot doesn't requre to calculate odometry data. We assume that robot can turn either left or right only 90degree. It can move only to adjacent state block not diagonal block. The planner takes average instantiation time 0.31secs and planning time 0.0177secs.