基于条件随机场和斯塔克尔伯格模型的决策系统研究

Research on Decision-making System Based on Conditional Random Fields and Stackelberg Models

文章依托驾驶模拟器平台,对CarSim、Simulink和VeriStand等软件进行设置,搭建硬件在环仿真平台,使用头部姿态估计网络算法检测头部姿态,采集试验数据。建立基于条件随机场的驾驶意图在线识别系统,对左换道、右换道和车道保持3种驾驶意图进行识别,与隐马尔可夫模型对比,证明该系统能够稳定输出驾驶员的驾驶意图。建立基于斯塔克尔伯格博弈模型的人机共驾决策系统,设立仿真试验场景,验证基于驾驶意图识别的人机共驾决策系统的有效性。

Based on the driving simulator platform,this paper establishes a hardware-in-the-loop simulation platform using software such as CarSim,Simulink,and Veristand.The head attitude estimation network algorithm is utilized to detect head attitude and collect experimental data.Furthermore,an online recognition system of driving intention based on conditional random field is developed to accurately identify three driving intentions:left lane change,right lane change,and lane keeping.The effectiveness of the system is demonstrated by comparing it with the hidden Markov model in terms of stable output of driver's driving intention.Additionally,a human-machine co-driving decision-making system based on Stackelberg game model is established and its effectiveness is verified through a simulation experiment scenario.