Robotic navigation based on logic-based planning

Abstract

Logic and Planning are interesting artificial intelligence problems in the context of robotic systems, i.e., robotic navigation. For such an autonomous system one of the requisites is that the goal has to be achieved without intervention of human being. We present a practical implementation of autonomous robotic navigation based on logic-based planning. We achieve this by using strength of PROLOG in order to generate plan to reach goal position from an initial. We utilize First Order Logic (FOL) that automatically asserts and retracts facts at runtime dynamically. All possible plans are computed using local search strategies (e.g., Depth and Breadth First) on state space representing a real, dynamic, and unpredictable environment. In order to navigate in the environment following optimized plan - one with fewest states, a balanced size 4-wheel differential drive robot has been carefully constructed. It can turn 90° and actuate forward by controlling linear (νt = 0.25m/s) and angular (ωt = Π/8 rad/s) velocities of two rear motorized wheels. It is also equipped with an Ultrasonic sensor to avoid collision with obstacles. The system is evaluated in an environment comprising of corridors with adjacent rooms. Graphical User Interface (GUI) is developed in .Net (C#) to map situation in Prolog and transmit plan to hardware for execution. Average time calculated for a plan to generate is 0.065 seconds. The robot moves block by block where each block in the state space represents 2m2 area. In addition to minors, our major contribution is that we offer a unified scheme for robotic navigation without calculating odometry data with the assumption the robot cannot be kidnapped nor slipped.