Research On Intelligence Level Of UAV Collective Intelligence System Based On Multi-Agent

Group or collective is a very common organizational structure in both nature and human society.In such an organizational structure,there are a variety of explicit or implicit interactions,so that the overall intelligence performance of the group cannot be simply regarded as the accumulation of multiple individuals.Collective intelligence is embodied when multiple agents act in a coordinated way.With the rapid development of information technology today,the application of unmanned aerial vehicles(UAV)is becoming more and more common.However,due to the relatively simple behavior and limited range of a single UAV,UAVs often perform tasks in groups,such as forming special icons for performance,monitoring and environmental reconnaissance,or serving as temporary mobile base stations in emergencies,etc.It is necessary to establish the corresponding quantitative evaluation mechanism before optimizing the intelligence level of the group.Based on the quantization of intelligence level of homogeneous groups,this study extends the Anytime Universal Intelligence Test(AUIT),which is based on grid space and information theory,to make it applicable to the quantization of intelligence level of heterogeneous groups.Then,this mechanism is applied to quantify the intelligence level of heterogeneous groups.According to the experimental results,the relationship between the intelligence level of heterogeneous groups and the related influencing factors is analyzed and summarized,and the importance of effective information interaction in the group to the intelligence level of heterogeneous groups is clarified.At the same time,the intelligence level of homogeneous groups and heterogeneous groups is compared,and the conclusion is that heterogeneity has a positive effect on the intelligence level of the groups.In the implementation of specific UAV collective intelligence system performance optimization,this study constructs a mission scenario where UAV groups are used as mobile base stations to cover ground users with communication signals.In order to enhance the effective information interaction in the group,a graph convolutional multiagent reinforcement learning algorithm,DGN,is applied to the UAV collective intelligence system.Experimental results show that this algorithm can provide collective with higher coverage degree,higher coverage fairness and lower UAV mobile energy consumption.This once again validates the positive impact of effective information interaction on the performance of collective intelligence.