高速公路施工区合流路段交通冲突模型

为了对高速公路施工区合流路段交通冲突进行微观分析以及探索交通冲突技术在无人驾驶领域的应用,采集12个交通微观信息变量数据作为训练集,结合先验知识,建立了贝叶斯网络模型并采用交叉验证方式评估模型精确度. 结果表明:大车作为冲突主体车辆会导致交通冲突概率和冲突严重性上升；与冲突两车的前后相邻车辆会对两车的冲突有一定影响；当车辆采取超车行为时,会提升车辆发生交通冲突的概率；匝道合流路段发生交通冲突的概率和严重性较转幅路段更高. 采用建立的贝叶斯网络交通冲突模型对驾驶策略进行评估并提出冲突规避措施,如分流大车；在特定路段采取管制措施禁止超车,限制单车道长度,设置车距确认设施等.

Traffic conflict model for confluence section in highway construction area

For the microscopic analysis of road traffic conflicts in confluence sections of highway construction area and to explore the application of traffic conflict technology in the field of unmanned driving, 12 variables of microscopic traffic information were collected as training set. The Bayesian network model was established combined with prior knowledge, and the precision of the model was evaluated by using cross validation method. Results show that large vehicles, as the main body of the conflict, could lead to the increase of the probability and severity of traffic conflicts. The preceding and following vehicles of two conflict vehicles had certain impact on the collision between the two vehicles. Overtaking behavior increased the probability of traffic conflicts. The probability and severity of traffic conflicts were higher in ramp merging sections than in turning sections. The proposed Bayesian network traffic conflict model was used to evaluate driving strategies, and conflict avoidance measures were put forward, such as diverging large vehicles. In certain road sections, it is suggested to take control measures to prohibit overtaking, limit the length of a single lane, set up distance confirmation facilities, and so on.