A review on motorcycle and rider modelling for steering control

The paper is a review of the state of knowledge and understanding of steering control in motorcycles and of the existing rider models. Motorcycles are well known to have specific instability characteristics, which can detrimentally affect the rider's control, and as such a suitable review of these characteristics is covered in the first instance. Next, early models which mostly treat riding as a regulatory task are considered. A rider applies control based on sensory information available to him/her, predominantly from visual perception of a target path. The review therefore extends to cover also the knowledge and research findings into aspects of road preview control. Here, some more emphasis is placed on recent applications of optimal control and model predictive control to the riding task and the motorcycle–rider interaction. The review concludes with some open questions which naturally present a scope for further study.     

电动二轮车驾驶人头部损伤再现不确定性方法

为了减少交通事故中不确定性信息在受伤害者损伤再现中造成的不利影响,采用拉丁超立方(LHS)试验设计和响应面蒙特卡洛相结合的方法对电动二轮车驾驶人头部损伤再现进行损伤不确定性分析。首先采用多体系统动力学方法对具有详细事故信息(含视频信息)和损伤记录的2起电动二轮车事故中的汽车碰撞速度进行再现和不确定性分析,并对比事故信息(视频信息、最终位置、电动二轮车驾驶人的运动学),进而验证事故再现结果的有效性。在此基础上,应用有限元方法将获得的电动二轮车驾驶人头部碰撞的边界条件加载至THUMS人体头部有限元模型,分析电动二轮车驾驶人头部损伤参数的不确定性与简明损伤准则AIS累计频率的分布关系,并对比电动二轮车事故案例中电动二轮车驾驶人的头部损伤法医鉴定记录。研究结果表明:蒙特卡洛不确定性分析方法能够较准确地预测电动二轮车事故中的汽车碰撞车速,采用该分析方法获得的电动二轮车驾驶人头部损伤等级与法医鉴定的脑部损伤记录高度吻合;蒙特卡洛不确定性分析方法可以适用于评估电动二轮车事故中电动二轮车驾驶人的头部损伤等级,研究结果可为电动二轮车驾驶人头部损伤研究提供理论依据和实证方法。     

基于差异性区域预测的行人与骑车人联合检测

本文中针对自动驾驶车辆在环境感知过程中易将行人与骑车人混淆的问题,提出一种有效区分行人与骑车人的联合检测方法,并基于快速区域卷积神经网络Faster R-CNN进行改进.首先,通过增加一个子网络提取图像形状特征通道,将其与主干网络生成的特征图进行聚合,额外的形状语义通道用以辅助检测器区分行人与骑车人的特征;接着,通过构...     

Optimal control to modelling motorcycle rider steering: local versus global coordinate systems in rider preview

The response of a motorcycle is heavily dependent on the rider’s control actions, and consequently a means of replicating the rider’s behaviour provides an important extension to motorcycle dynamics. The primary objective here is to develop effective path-following simulations and to understand how riders control motorcycles. Optimal control theory is applied to the tracking of roadway by a motorcycle, using a non-linear motorcycle model operating in free control by steering torque input. A path-following controller with road preview is designed by minimising tracking errors and control effort. Tight controls with high weightings on performance and loose controls with high weightings on control power are defined. Special attention is paid to the modelling of multipoint preview in local and global coordinate systems. The controller model is simulated over a standard single lane-change manoeuvre. It is argued that the local coordinates point of view is more representative of the way that a human rider operates and interprets information. The simulations suggest that for accurate path following, using optimal control, the problem must be solved by the local coordinates approach in order to achieve accurate results with short preview horizons. Furthermore, some weaknesses of the optimal control approach are highlighted here.