Escape from local minima for virtual agent navigation in unknown virtual environments

This paper presents a solution for escaping from local minima using fuzzy logic. The problem occurs when a virtual agent must navigate to pass an obstacle on heading towards a desired target with no prior knowledge of the environment and gets trapped in a dead-end. We introduce a behavioural-based architecture which comprises of goal- seeking, obstacle-avoidance and path-planning behaviours, where each behaviour operates independently. The path-planning behaviour can, in real-time, identify local minima without the need for learning. We use the fuzzy alpha-level technique as a behaviour selection method to decide which behaviour task needs to be executed when the local minima problem occurs. The evaluation results demonstrate that a virtual agent can successfully escape from being trapped in local minima in various virtual environments.