Co-evolutionary path optimization by ripple-spreading algorithm

Abstract

Inspired by the multi-agent co-evolving nature reflected in many methods of evolutionary computation, this paper proposes a challenging routing problem — co-evolutionary path optimization (CEPO). Static path optimization (SPO) is a foundation of computational intelligence, but in reality, the routing environment is usually time-varying (e.g., moving obstacles, spreading disasters and uncertainties), and therefore dynamic path optimization (DPO) has to be addressed. To resolve DPO, a common practice is as following: at each time t, environmental parameters are measured/predicted first, and then the best path is re-calculated by resolving SPO based on the newly measured/predicted environmental parameters. In other words, during the path optimization process of time t, the routing environment is actually fixed and static. Usually, DPO cannot lead to optimal actual travelling trajectory, and it demands high online computation capacity. Distinguishing from DPO, in CEPO, future environmental parameters keep changing during the optimization process of time t. In other words, the routing environment co-evolves within the path optimization process of time t. No existing method can address CEPO without losing optimality or efficiency. This paper then reports a ripple-spreading algorithm (RSA), which can resolve CEPO with both optimality and efficiency. Surprisingly, in just a single offline run of RSA for the CEPO, the optimal actual travelling trajectory can be achieved in a given dynamical routing environment. Simulation results clearly demonstrate that: (i) the solutions to CEPO are far better than those to DPO, and (ii) the reported RSA is an effective and efficient method for addressing CEPO.