Intelligent Attack Behavior Portrait for Path Planning of Unmanned Vehicles

Abstract

With the rapid development of artificial intelligence, opponents can use AI technology to influence the path planning algorithm of unmanned vehicles, making unmanned vehicles face severe safety issues. Aiming at the opponent’s intelligent attack in the scenario of unmanned vehicle path planning, this paper studies the opponent’s intelligent attack behavior portrait technique and proposes an attack behavior portrait scheme based on the knowledge graph. First, according to the simulation experiment of unmanned vehicle path planning based on reinforcement learning, we use Toeplitz Inverse Covariance-based Clustering (TICC) time-series segmentation clustering technology to extract the steps of an opponent’s attack behavior. Then, the attack strategy rules are stored in the knowledge graph to form a portrait of attack behavior for unmanned vehicle path planning. We verified the proposed scheme on the Neo4j platform. The results proved that the method could describe the steps of intelligent attacks on unmanned vehicles well and provide a basis for unmanned vehicle attack detection and establishing an unmanned vehicle defense system. Furthermore, it has good generalizability.