Collision-free path finding for dynamic gaming and real time robot navigation

Abstract

Collision-free path finding is crucial for multi-agent traversing environments like gaming systems. An efficient and accurate technique is proposed for avoiding collisions with potential obstacles in virtual and real time environments. Potential field is a coherent technique but it eventuates with various problems like static map usage and pre-calculated potential field map of the environment. It is unsuitable for dynamically changing or unknown environments. Agents can get stuck inside a local minima incompetent in escaping without a workaround implementation. This paper presents efficient and accurate solutions to find collision free path using potential field for dynamic gaming and real time robot navigation. A surfing game in two testing environments with a Gamecar and a physical robot called Robocar is created with dynamic and solid obstacles. Sensor like proximity, line and ultrasonic are used along with the camera as different agents for path finding. The proposed intelligent agent (IA) technique is compared with other path planing algorithms and games in terms of time complexity, cost metrics, decision making complexity, action repertoire, interagent communication, reactivity and temporally continuous. It traverses for 135 meters(m) in 55.8 seconds(s) covering 20 goals and 419.3 m in 8.7 minutes while avoiding 10 local minimas successfully. Proposed technique shows comparable results to path finding with techniques using neural networks and A* algorithm. Experimental results prove the efficiency with run time overload, time complexity and resource consumption of the proposed technique.