Dolphin-inspired adaptive coordination for online multi-robot path planning in dynamic environments

Introduction Online multi-robot path planning requires repeated joint motion decisions amid static obstacles, moving obstacles, and other robots. Methods This study introduces the Dolphin Cooperative Foraging Optimizer (DCFO), a population-based method that adapts its search strategy within each replanning call. DCFO combines an elite-guided operator with three differential evolution operators. Historical operator-feedback counts determine selection probabilities; a stage-dependent rule determines which selected operator is executed; and population size decreases with the nominal evaluation budget. A masked, regularized inverse covariance perturbation is triggered when all population fitness values coincide. Search performance was evaluated on the CEC2020 and CEC2022 benchmark suites, followed by MATLAB simulations with 6 and 12 robots. Results Compared with four population-based planners and the Dynamic Window Approach, DCFO achieved the shortest aggregate path in both tested scenarios. Path length was 2.74% shorter than the next-best result in the 6-robot scenario and 8.30% shorter in the 12-robot scenario. DCFO also recorded the lowest average simulation execution time in the 12-robot scenario, although it was not the fastest method in the 6-robot scenario. Discussion These results identify adaptive operator portfolios as a promising approach to joint motion-command generation in dynamic multi-robot simulations. Future research will extend the framework to physical robot platforms and more diverse operating conditions.