No abstract

Automated vehicles (AVs) are deemed to be the key element for the intelligent transportation system in the future. Many studies have been made to improve the AVs' ability of environment recognition and vehicle control, while the attention paid to decision making is not enough though the decision algorithms so far are very preliminary. Therefore, a framework of the decision-making training and learning is put forward in this paper. It consists of two parts: the deep reinforcement learning (DRL) training program and the high-fidelity virtual simulation environment. Then the basic microscopic behavior, car-following (CF), is trained within this framework. In addition, theoretical analysis and experiments were conducted on setting reward function for accelerating training using DRL. The results show that on the premise of driving comfort, the efficiency of the trained AV increases 7.9% compared to the classical traffic model, intelligent driver model (IDM). Later on, on a more complex three-lane section, we trained the integrated model combines both CF and lane-changing (LC) behavior, the average speed further grows 2.4%. It indicates that our framework is effective for AV's decision-making learning.

Turning behaviour is one of the most challenging driving manoeuvres that take place at intersections. Autonomous vehicles (AVs) are often overly conservative in these scenarios as they compromised to others’ behaviour to realize behavioural consistency. This paper proposes an intended cooperative motion planning (ICMP) that can actively predict the intentions of interacting participants. This helps achieve socially compliant cooperation and enables each vehicle to converge upon a set of consistent behaviours. The ICMP framework divides the integrated driving process into two related modules: prediction and planning. These modules are integrated using the mechanism of intended cooperation, which first introduces the criteria for a cooperative response and then maps them in the motion planning space. The prediction module applies a Gaussian mixture model to learn human drivers’ behaviour to determine conflict points, which helps to narrow down the solution spaces. In the planning module, paths are represented by quartic Bézier curves and speed is modelled as a polynomial function. Optimizations can then be used that maximize the efficiency and smoothness for path planning, and comfort and efficiency for speed planning. The test results show that ICMP‐based AVs had consistent interactions with other vehicles. Moreover, when compared with a potential field‐based method, the ICMP‐based method had better performance in terms of safety, comfort and efficiency, especially when interacting with multiple oncoming vehicles.