SVM based Intention Inference and Motion Planning at Uncontrolled Intersection

This paper presents a support vector machine (SVM) based intention inference and motion planning algorithm for autonomous driving through uncontrolled intersection. Intention of target vehicles is inferred using SVM with intersection map to predict the future state of targets. A cross point, which has a highest collision probability, is estimated using predicted target state considering prediction uncertainty. Longitudinal acceleration is determined using model predictive control approach considering the predicted cross point. The proposed algorithm is validated via simulation and vehicle tests. The results show the accurate intention inference and human-like motion planning at uncontrolled intersection scenarios.