An online approach for intersection navigation of autonomous vehicle

Navigation through an intersection is a fundamental task that will enable an autonomous car to operate in a real traffic environment. Previous studies about intersection navigation generally assume vehicle to vehicle communication ability for all of the vehicles. Since this is unattainable in the near future, we focus on the scenario that vehicles on the road cannot communicate with each other. A new model is presented for this kind of intersection navigation as a Partially Observable Markov Decision Process problem. The proposed model can handle multiple numbers of cars in a dynamic environment. To validate the feasibility of the model, experiments are carried out with an autonomous golf cart in the university campus.