A Diffusion-Based Reactive Approach to Road Network Cooperative Persistent Surveillance

This paper addresses the problem of road network cooperative persistent surveillance algorithm suitable for online planning. A diffusion-based reactive cooperative path planing approach (DRCP) is designed. Compared with existing related methods, DRCP uses reactive architecture, probability independent decision strategy, and implicit cooperation technique, which achieve the fast computation, stable result, and accurate response. The characteristics of DRCP is to create a new exclusive cognition model for each unmanned ground vehicle (UGV), which allows each UGV to refer to only one virtual variable to achieve: 1) Global state estimation based on a limited cognition space. 2) Avoidance of local optimum. 3) Global cooperation with coordinators. Theoretical analysis proves that DRCP may quickly plan paths for a UGV group with any sensor models meeting specific conditions. Numerical simulation shows the accuracy and stability of DRCP. The robustness to unstructured disturbances and rapid response demonstrated in outdoor experiment indicate the ability/potential of the approach for online planning.