Navigation patterns of a hybrid scanning agent using uninformed and informed search algorithms for reactive and dehberative behaviors

Abstract

Hybrid agent architectures enable effective control layering the agent functions according to the sophistication they require. In this study, assuming a partially observable and deterministic environment, reactive and dehberative behaviors of a hybrid scanning agent were controlled using some uninformed and informed search algorithms, respectively. As the reactive layer decided the agent actions online in an unknown environment, a map of the environment was constructed in parallel. When the reactive layer failed to find a proper action, the deliberative layer proposed a solution offline using the constructed map so that the agent could continue its scanning task. The navigation patterns that were produced with the adapted algorithms in the reactive and dehberative layers were analyzed. The results showed that depth-first search (DFS) and breadth-flrst search (BFS) algorithms can be used as reactive motion planners for scanning an environment in zigzag and spiral patterns. Simulations in 25 grid-based environments with different sizes and varying percentages of obstacles yielded that running A* algorithm as a dehberative planner, the agent could completely scan the environments with equal successes for all different modes of the developed scanning algorithms. The horizontal mode of the DFS-based scanning algorithm had the least rescan rate on average.