决策系统是无人驾驶技术的核心研究之一。目的：已有决策系统存在逻辑不合理、计算效率低、应用场景局限等欠缺，因此提出一种动态环境下无人驾驶路径决策仿真。方法：首先，基于规则模型构建适于无人驾驶决策系统的交通有限状态机；其次，针对交通动态特征，提出基于统计模型的动态目标路径算法计算状态迁移风险；最后将交通状态机和动态目标路径算法有机结合，设计出一种基于有限状态机的无人驾驶动态目标路径模型，适用于交叉口冲突避免和三车道换道行为。将全速度差连续跟驰模型运用到换道规则中，并基于冲突时间提出动态临界跟车距离。结果：为验证模型的有效性和高效性，对交通环境进行虚拟现实建模，模拟交叉口通行和三车道换道行为，分析文中模型对车流量和换道率的影响。实验结果显示，在交叉口通行时，自主车辆不仅可以检测冲突还可以根据风险评估结果执行安全合理的决策。三车道换道时，自主车辆既可以实现紧急让道，也可以通过执行换道达成自身驾驶期望。通过将实测数据和其他两种方法对比，当车流密度在0.2到0.5之间时，本文模型的平均速度最高分别提高32 km/h和22 km/h。当车流密度不超过0.65时，本文模型的换道成功率最高分别提升37%和25%。[将精度写在了结果中]结论：实验结果说明本文方法不仅可以在动态城区环境下提高决策安全性和正确性，还可以提高车流量饱和度，缓解交通堵塞。
Objective Driverless technology is an essential part of the intelligent transportation system, including environmental information perception, intelligent planning and multi-level auxiliary driving. It reduces driver''s work intensification and prevent accidents. With the development of artificial intelligence, autonomous vehicle has become a hot topic in both industry and academia in recent years. In addition, decision making system is one of the core research of driverless technology. Reducing the number of road accidents is of paramount societal importance, and a growing research effort has been devoted to decision-making systems within the last few years. Making human-like decisions with other encountering vehicles in complex traffic scenarios brings big challenges to autonomous vehicles. The research of the autonomous driving decision system has important theoretical and practical value for improving the level of intelligent vehicle and intelligent transportation systems. However, due to the uncertainty and randomness of the driving behavior of the surrounding vehicles, the current decision-making system has some shortcomings, such as unreasonable logic, large computational complexity and limited application scene. In order to solve these problems, the purpose of this paper is constructing a finite state machine based decision-making system for autonomous vehicles are safely driven in the dynamic urban traffic environment. This paper mainly studies how vehicles pass through intersections and how to change lanes, which are the core issue of decision-making system. Method In a certain period of time, the driver''s behavior is determined by the current traffic condition and risk perception. We define the primary state of the vehicle according to the driving range of the autonomous vehicle, including driving at the intersection, driving in the driveway, and approaching to the crossroads. Each primary state includes many secondary states. For example, at crossroads, the vehicle may turn or keep straight. Combining the original finite state machine theory, a traffic state machine (TSM) suitable for intelligent systems is proposed. Considering the complexity and diversity of the traffic environment, a dynamic target path (DTP) algorithm is proposed to improve the feasibility of the decision system. combining the traffic state machine with the dynamic target path algorithm, we propose the dynamic target path model based on finite state machine applied to the decision system, and analyze the significance of the model. In complex and diverse traffic environment, intelligent vehicles only pay attention to their own driving information, ignoring the state of other vehicles, which brings great risks. As a result, we divide the awareness area and the conflict area for each autonomous vehicle. At the crossroads, the perceived range of autonomous vehicles is defined as the awareness area, and the reachable range of autonomous vehicles is called the conflict area. In the driveway, the perception area of vehicle is defined as the consciousness area, and the range of interaction between autonomous vehicles and surrounding vehicles is defined as the conflict area. A reasonable decision can effectively reduce the probability of accidents in conflict areas. We use dynamic target path algorithm (DTP) to calculate the risk of decision-making to restrict vehicle behavior. Due to a fixed follow-up distance has no ability to consider the influence of the speed, this paper proposes a dynamic critical follow-up distance, which reduces the collision with the preceding vehicle while following the vehicle. Furthermore, the Full Velocity Difference model is used to avoid collision with the front vehicle of the target lane during the lane change under different scenarios. Results In order to verify the effectiveness of the model and algorithm, we perform experiments in different scenarios repeatedly through the Unity 3D engine. In the first experiment, we simulate the scene of autonomous vehicle driven at a crossroad. The second experiment simulates how the autonomous vehicles respond to emergencies. The third experiment simulates that autonomous vehicles change lanes to reach their destinations. The fourth experiment simulates that autonomous vehicles to change lanes to improve their speed. In order to prove that the experimental results are not affected by the bad weather, this paper simulates the lane changing behavior of autonomous vehicles in foggy days. Experiments show that in bad weather, autonomous vehicles can not only meet the driving expectation, but also ensure driving safety. Experimental results show that, on the basis of other vehicles driving intentions can be obtained, and autonomous vehicles has the capability to make correct decisions according to the potential risk of intersection and current traffic environment. And autonomous vehicles can change lanes according to their driving demand when driving on the driveway. In case of emergency, the autonomous vehicle regards the special vehicle as a dynamic obstacle. After yielding the right-of-way to emergency vehicles, the autonomous vehicle returns to the original lane to continue driving. In order to prove that the proposed method can improve the efficiency of traffic flow, and compare the model proposed in this paper with other models. Results demonstrate the difference among the three models is not obvious when the vehicle density is small. However, when the vehicle density is greater than 0.2 and less than 0.5, the average speed of the model is increased at most by 32 km/h and 22 km/h respectively. When the density of vehicles is not more than 0.65, the success rate of lane changing in this model is increased at most by 37 percentage points and 25 percentage points respectively. Conclusions The algorithm in this paper not only improves the safety and accuracy of decision-making in the dynamic urban traffic environment, but also helps improve traffic flow saturation and reduce traffic flow. In addition, a variety of traffic environments can be modeled by our simulation framework. Although the proposed model and algorithm are relatively simple, the assessment of potential risks can meet the planning time of autonomous driving. Our work sets down rules for the decision-making of autonomous driving, and provides some references for the development of the intelligent transportation system. However, the influence of vehicle types, trajectory and road width on decision-making are neglected. In the future, we will improve the current work and provide a more complete and reasonable framework for the automatic driving decision system.