一种自动泊车的仿人智能控制方法

Autonomous Parking Method Based on Human-simulated Intelligent Control

为简化平行泊车,通过总结模拟熟练驾驶员的泊车经验,提出一种自动泊车的仿人智能控制方法。该方法以后轮驱动、前轮转向的四轮汽车为对象,建立车辆运动学模型;通过分析熟练驾驶员泊车流程,将泊车过程分为4个阶段;在相切圆弧加公切线的规划路径上,选取泊车阶段转换时车辆姿态调整的关键点作为跟踪目标;根据泊车过程车辆的位姿信息,提取12种泊车的特征状态,作为描述车辆泊车动态行为的特征模型;根据熟练驾驶员的泊车策略,构建控制模态集。泊车开始后,依据特征模型先验知识和当前车辆位姿与泊车目标的偏差,对当前车辆特征状态进行模式识别,由辨识出的特征状态驱动相应控制模态,控制汽车按规划路径泊车入位。建立了车辆运动学和仿人智能控制器Simulink模型,并进行了仿真实验。仿真结果表明,该方法能有效控制车辆泊入车位。

To simplify parallel parking, an autonomous parking method based on human-simulated intelligent control is proposed, by summarizing and imitating the parking skills of experienced drivers. A four-wheeled vehicle with rear-wheel drive and front wheel steer- ing is used as study object. The vehicle kinematic model is build. According to the drivers＇ experience, the parking processing is divid- ed into four stages. Four key points on the programming trajectory composed with two tangent arcs and a tangent are chosen as tracking targets. At these points, conversion between two consecutive parking stages takes place and the vehicle pose is pivotal adjusted. Twelve character modes are extracted to describe dynamic behaviors of the parking feature from the information of the vehicle pose. Control mode set is developed according to the parking strategy of skilled drivers. When the parking begins, pattern recognition is used to iden- tify the characteristics model according to the a priori knowledge of character models and the deviation of vehicle position. To control the vehicle parking along the programming trajectory, corresponding control mode is adapted by the characteristics model. Simulink kine- matic modes of the vehicle and the proposed human-simulated intelligent controller are established. Simulations had been done and the results demonstrated the effectiveness of the proposed method.