A Cooperative Control Strategy for Predicting Passing Capacity and Intelligent Obstacle Avoidance in Autonomous Vehicles Based on Multisensor Fusion

The prediction of a vehicle’s ability to traverse specific roads with rough terrains or obstacles and the passing ability of autonomous vehicles remains largely unexplored within current research. Therefore, to predict the passing performance of autonomous vehicles on structured roads, a collaborative control method for vehicle obstacle avoidance and passing performance was established. First, a joint calibration and information acquisition system using millimeterWave radar (mmWave radar) and cameras was developed to address small and medium-sized obstacles or trenches on the road that affect the passing ability of vehicles. Subsequently, the geometric bounding-box model parameters of the target obstacle were extracted, and the bounding-box feature value method was used to match and fuze data, thereby generating highly reliable obstacle feature information. A mathematical model for vehicle passing ability was established to derive driving limit data, clarifying the obstacles that a vehicle can navigate through. Based on the location and geometric feature information of obstacles and passing ability limit parameters of the vehicle, the passing ability of unmanned vehicles was predicted, and reasonable traffic and obstacle avoidance control strategies were proposed. Moreover, vehicle passing ability tests were conducted on structured road surfaces. The experimental results demonstrated that at a speed of 30 km/h, the success rates of the model vehicle in traversing obstacles and steps were 97.4% and 96.6%, respectively. These outcomes highlighted the feasibility of vehicles traversing vertical obstacles and trenches. The proposed approach can enhance the safety of intelligent vehicle operations.