Task Allocation for Gantry-Type Multi-Robot Systems with Heterogeneous Welding Tasks

To address the task allocation problem of a gantry-type dual-arm welding robotic system subject to heavy and diverse welding workloads as well as limited operating span, a Dynamic Small-World Genetic Algorithm (DSWGA) for multi-objective optimization is proposed to obtain optimal welding task allocation. To meet the operational requirements of workload balancing and high-efficiency welding for ship multi-subassembly components, a multi-objective optimization model is established to minimize the system makespan and the load imbalance rate, and the small-world genetic algorithm is adopted to solve this model. To tackle the drawback of fixed parameters in the small-world network, a dynamic network parameter adjustment mechanism is developed. By dynamically tuning the connectivity among individuals in the population, the scope of information exchange is adjusted throughout the optimization process to enhance the global search capability of the algorithm. To overcome the limitation of fixed crossover and mutation probabilities in conventional genetic operations, adaptive crossover and mutation strategies are designed based on the structural differences between parents and the structural similarity between offspring and non-dominated solutions, which improves the search efficiency and solution-set quality. Test results for each improvement strategy demonstrate that the three strategies show potential complementary effects, and their combination yields better overall performance in terms of the mean values of HV, GD, and Spread, with a statistically significant advantage in HV. Task allocation experiments are conducted in three typical welding scenarios with different task scales and constraint characteristics. The results indicate that DSWGA achieves competitive overall performance in convergence, solution set quality, and distribution uniformity compared with NSGA-II, SPEA2, and MOPSO. In the more complex test scenario, the makespan of the selected compromise solution obtained by DSWGA is reduced by approximately 1.03%, 3.72%, and 2.68% compared with NSGA-II, SPEA2, and MOPSO, respectively, while the robot load balancing rate reaches 99.7%.