Research on the multiple unmanned aerial vehicle swarm confrontation strategy based on the dynamic Bayesian network

Abstract

The swarming confrontation problem of unmanned aerial vehicles (UAVs) has been a focal point and challenge in the field of complex systems research in recent years, with significant application value in the military, network security, and artificial intelligence industries. In real-world confrontations, the uncertainty of the environment and the diversity of intelligent agent behaviors render the problem difficult to model, and operational environments necessitate intelligent agents to provide real-time, efficient responses to changes in the situation. To address these challenges, this paper proposes a research framework for the swarming confrontation problem of UAVs. First, an adversarial game model is developed based on the improved Lanchester equation to solve this problem toward the red and blue swarms. The adversarial game model focuses on describing the dynamic quantity change process of the red and blue UAV swarms. Second, based on the above confrontation model, a multiple-task assignment problem is investigated, which is derived from the above confrontation process. A new assignment strategy of these UAVs for strike tasks is proposed by adaptive improvement based on the traditional Kuhn–Munkres algorithm. This strategy is suitable for the red and blue parties under the adversarial environment, which can effectively complete the strike tasks and improve the confrontation ability of these UAVs. Third, a swarming confrontation algorithm is proposed to improve the environmental suitability of each UAV, especially when dealing with the influences of a series of uncertain factors generated by the real-time process of swarming offensive and defensive confrontations. This algorithm is based on the dynamic Bayesian network structure and focuses on predicting and evaluating the uncertainty generated during the confrontation process of the red and blue UAV swarms and performing corresponding reasoning and prediction through the dynamic Bayesian network, which can effectively reduce the complexity and calculation of the confrontation model and widely improve the accuracy and speed of decision-making. Finally, a Python-based real-time simulation platform is built on the above-mentioned confrontation model to illustrate the evolutionary process of the red and blue UAV swarms and to verify the effectiveness of the proposed algorithm by comparison with the most classic artificial potential field method. The simulation results reveal that the above framework can demonstrate the real-time offensive and defensive confrontation process of red and blue UAV swarms according to the designed penetration mission, effectively solve the problem of task assignment conflicts between red and blue swarms, properly predict and evaluate the uncertainty issues derived from the swarm confrontation process, and improve the combat capabilities of UAV swarms.

Keywords

• swarm confrontation
• distributed cooperation
• lanchester equation
• kuhn–munkres algorithm
• dynamic bayesian network