A positioning system estimates the position and orientation of a vehicle. Autonomous driving systems plan paths and control the vehicle based on the information from the positioning system. Recently, methods of estimating the position in urban areas have been actively studied. In U-turn sections, which are common in urban areas, vehicles perform a rotation to reverse the direction of travel. Through these sections, drivers can reduce the travel distance and save time but with a high risk of an accident. Despite there being a need for the development of autonomous driving schemes for U-turn sections, the existing research is limited. This study proposes an interacting multiple model (IMM) filter-based position estimation algorithm for urban U-turn sections. To reflect the dynamic characteristics of a vehicle during U-turn maneuvers, a multiple vehicle model was used. This model includes kinematic and dynamic vehicle models. The state estimates of the vehicle model and gyroscope are combined using an IMM filter. The position estimation algorithm developed in this study is verified via experiments. The experimental results indicate that, during urban U-turn maneuvers, the position estimation accuracy of the IMM filter-based algorithm is improved than that of the single vehicle model.

定位系统估计车辆的位置和方向。自动驾驶系统根据来自定位系统的信息规划路径并控制车辆。近来，已积极研究估计城市区域中的位置的方法。在市区常见的掉头路段，车辆通过旋转来反转行驶方向。通过这些路段，司机可以减少行驶距离并节省时间，但发生事故的风险很高。尽管需要为掉头路段开发自动驾驶方案，但现有的研究是有限的。本研究提出了一种用于城市掉头部分的基于交互多模型 (IMM) 滤波器的位置估计算法。为了反映车辆在掉头机动过程中的动态特性，使用了多车辆模型。该模型包括运动学和动力学车辆模型。车辆模型和陀螺仪的状态估计使用 IMM 滤波器进行组合。通过实验验证了本研究中开发的位置估计算法。实验结果表明，在城市掉头机动过程中，基于IMM滤波器的算法的位置估计精度比单车模型的位置估计精度有所提高。