AI-Empowered Task Intent Prediction and Path Planning for Cooperative AAV Swarms in Consumer Internet of Vehicles

Abstract

The integration of Autonomous aerial vehicles (AAV) significantly expands the operational scope of the Consumer Internet of Vehicles (CIoV), enhancing its potential applications. In this paper, a AAV swarm task intent prediction and path recommendation framework is proposed to effectively address the challenges of resource utilization associated with spatially distributed consumer electronic services. Employing fuzzy theory, the framework integrates multi-sensor AAV data using Neural Network (NN)-based membership functions and fuzzy rules. These elements facilitate automatic optimization and historical scenario analysis to infer task intent, represented as spatial task distribution. Additionally, the predicted intent and environmental features are combined in a Long Short-Term Memory (LSTM) network to predict bid prices in an auction algorithm, thereby enhancing path recommendation capabilities. The framework is validated on the developed semi-physical platform supporting real-time AAV swarm control and inter-AAV communication, demonstrating superior performance in system throughput and service percentage compared to existing methods.