Determination of Weight Coefficients for Multi-Objective UAV Task Assignment via Inverse Optimization and Adaptive Adjustment of Planning Intent

This paper presents a method for determining weight coefficients in multi-objective UAV task assignment under a competitive data-collection setting, where two UAV companies independently visit each other’s sensor nodes. In this setting, each company expresses operational intent through weights for data value, distance utility, and flight safety. Our own weights are prescribed by higher-level planning, whereas the competitor’s latent weights must be estimated from observed visit assignments. To estimate these weights, we apply Approximate Bayesian-Inspired Inference, which also provides a measure of the associated inference uncertainty. Subsequently, a purpose-built fuzzy rule base translates the inferred weights into response weights, while the fuzziness of this mapping is quantified as a second source of uncertainty. From these two uncertainty measures, an adaptive coefficient is derived that blends the planning and response weights into the final task assignment. Under controlled synthetic settings, numerical experiments demonstrate that the inference procedure outperforms direct loss-minimization baselines in both accuracy and speed, and that the adaptive fusion effectively balances planning adherence with responsiveness to competitor behavior.