Multi-agent path planning approach using assignment strategy variations in pursuit of moving targets

This study aims to investigate the problem of assignment strategies for multiple agents. In multi-agent scenarios, agents compute a path towards the goal, while these goal destinations in some cases are predefined in advance. The topic of assignment strategies, where agents need to identify and assign goal destination at the initial position, before making any move, has not been studied sufficiently. This problem becomes even more challenging if the goal destinations change over the period of time. This study investigates for new approaches to the assignment strategy in order to improve the efficiencies introducing three novel approaches for multiple agents in multiple moving targets environments: twin-cost, cover-cost and weighted-cost criteria. These new methods have been tested against existing overall the best approach in the literature. Empirical analysis is performed on grid-based gaming benchmarks. The performance is measured for the successful completeness of the test runs and achieving the shortest distance travelled. The experimental results suggest that the new assignment strategy methods exhibit better results in comparison with the existing approaches, where even some individual cases improve approximately by 23% especially when the means are the same.