Tag
This paper proposes a reward-density heuristic for dynamic multi-vehicle routing problems that matches the solution quality of metaheuristics like ALNS, GA, and SA while requiring two to three orders of magnitude less planning time, demonstrated on drone task allocation and urban taxi dispatch scenarios.
Introduces RACL, a reasoning-agent control layer that improves metaheuristic optimization by learning to control internal search behavior from operational memory, showing cost improvements in vehicle routing tests.