智能汽车在提高行驶安全性和减少交通事故方面有很大的优势，已成为世界范围内的研究热点。综述了智能汽车横向控制的国内外发展历程与研究现状；介绍了车辆横向动力学和轮胎力学的研究历程和模型；阐述了智能汽车横向控制理论和方法以及自动转向执行机构的设计；给出智能汽车横向控制研究的重点和发展趋势。通过分析认为，系统非线性、不确定性和时变特性的智能汽车横向动力学建模和横向控制器设计，特别是高速时的横向控制，以及智能车辆感知决策系统与车辆本身系统的一体化设计，将是今后研究的重点。

越野工况履带车辆动力学是车辆越野行驶能力预测的理论基础。考虑高机动履带车辆越野行驶纵向运动和垂向运动耦合效应，提出履带车辆纵向与垂向耦合动力学建模方法。建立车辆耦合动力学模型，对车辆在典型路面上的行驶性能进行仿真，并进行实车测试验证，进而量化分析履带车辆系统的功率特性。仿真结果表明：该模型可以表征车辆在不同类型路面上行驶的动力学响应特性；在越野工况下，车体垂向、俯仰等运动将消耗部分功率，对车辆行驶速度提升有一定影响；在越野路面上行驶的履带车辆瞬时非纵向运动功率数值波动范围大，随着路面条件变差以及速度提高，非纵向运动功率占总输入功率比例增加，对总功率的需求也越大，限制了驱动功率的有效利用。研究结果可用于高机动履带车辆动力学模型构建以及车辆功率流分析。

The dynamics of tracked vehicle in the off-road condition is the basis of the prediction of the off-road driving ability. A longitudinal and vertical coupled dynamic modeling method of tracked vehicle is presented，in which the coupling effect of longitudinal and vertical motion of high speed tracked vehicle is considered. A vehicle coupling dynamic model is established. The driving performance of vehicle on the typical road was simulated and verified in the real vehicle test. On this basis，the power flow characteristics of the tracked vehicle were analyzed. The simulated results indicate that the proposed model can be used to calculate the influence of the different road conditions on the vehicle driving performance. In the off-road condition，the road roughness makes the body vertical，pitching and other movements consume part of the power，and then have a certain impact on the speed of the vehicle. The instantaneous non-driving power of the tracked vehicle fluctuates widely. With the deterioration in the road conditions and the increase in speed，the ratio of non-driving power to total input power increases.The demand for the total power increases，which limits the effective use of the driving power.The results can be used for the dynamic model construction and the power flow analysis of high-speed tracked vehicles.

A motion planning method based on the offline generation and online extension and selection of motion primitives is proposed for planning the motion of heterogeneous tracked vehicles under a unified algorithm framework. The heterogeneous tracked vehicle studied refers to a tracked vehicle equipped with three different steering mechanisms. The optimized generation of motion primitives takes trajectory smoothness as the goal. A motion primitive library that meets the motion planning requirements of heterogeneous tracked vehicles is generated by comprehensively considering the differentiated behavior constraints of heterogeneous tracked vehicles, platform kinematics model constraints, smooth transition constraints, etc. The extension and selection of motion primitives are based on a unified algorithm framework, which comprehensively considers the attributes of the motion primitives, reference lines, expected velocity, obstacles, and other factors. By selecting the appropriate primitives from the motion primitive library to form the primitive sequences, the coordinated motion at the level of trajectory and velocity is planned. The results show that the proposed unified motion planning method for heterogeneous tracked vehicles, with the help of differentiated behavior constraints when primitives are generated, and unequal selection weights set according to primitive attributes, generates a motion planning result that reflects the difference in characteristics of heterogeneous tracked vehicles and meets real-time requirements.

针对重载子午轮胎扁平率接近1的特点，建立面内刚柔耦合模型并开展振动分析方法研究。提出重载轮胎面内胎体与胎侧耦合的试验模态测试与分析方法；考虑胎侧的惯性力和分段刚度，建立重载轮胎面内弹性基础柔性梁动力学模型；利用有限差分法，将面内耦合动力学方程离散化，建立基于轮胎几何、结构参数的三参数等效刚度重载轮胎面内刚柔耦合模型；基于试验和解析模态共振频率，利用遗传算法对重载轮胎结构参数辨识，建立重载轮胎面内刚柔耦合模型的传递函数解析模型。数值分析和试验验证结果表明：在0~300 Hz频率范围内，面内刚柔耦合模型可准确表征重载轮胎面内胎体和胎侧的耦合振动模态与传递特性；该建模与试验分析方法可将重载轮胎的分析频率由0~180 Hz扩展至300 Hz.

轮地相互作用力对于车辆设计、通过性评价、控制和仿真等方面具有极其重要的作用。传统模型以平板压力为力学假设前提，导致模型中各项参数难以同时响应轮胎几何和胎压因素对轮地相互作用力的影响。采用特定几何和胎压条件下的轮胎弹性形变量建立虚拟刚性轮形，结合轮地接触应力的分布状况，推导一种新的轮地相互力模型。该模型依据轮胎运动状态划分为轮胎准静态压载的垂向载荷模型以及轮胎稳态滑转的牵引力模型，模型中各项参数具有反映轮胎几何、胎压和土壤力学性质之间关联性的特点。通过土槽和原位地面试验对轮地相互作用力模型及其参数辨识结果进行验证，试验结果表明，不同几何尺寸和胎压下的轮胎垂向载荷与试验值之间最大误差不超过0.1 kN，模型参数辨识结果与试验值之间相对误差不超过12%，依据参数辨识结果计算的轮胎牵引力与试验值之间方均根误差为0.37。因而该模型可以有效地应用于考虑轮胎几何和胎压因素的轮地相互作用力学计算中。

Tire-ground interaction plays an important role in vehicle design, trafficability evaluation, control and simulation. The traditional model is based on the assumption of plate pressure, which makes it difficult for the parameters of the model to simultaneously respond to the influence of tire geometry and tire pressure. The virtual rigid wheel shape is established by using the elastic deformation of tire under specific geometry and pressure, and based on the stress distribution of rigid wheel shape, a new tire-ground interaction model is derived. According to tire motion state, the model can be divided into vertical load model with tire quasi-static load and pulling force model with tire steady slip, which can reflect the correlation between tire geometry, tire pressure and soil mechanical properties. The model of tire-ground interaction force and its parameter identification results are validated by soil-bin and in-situ tests. The test results show that the maximum error between calculation and test is not more than 0.1 kN under different tire geometric sizes and tire pressures, and the relative error between the parameter identification result and the test result is less than 12%. The root mean square error between the tire tractive force and the test result is 0.37. Therefore, the model can be effectively applied to the mechanical calculation of tire-ground interaction considering tire geometry and tire pressure.

胎压的变化会显著地影响轮胎力学特性，进而影响车辆的操纵稳定性，而当前关于胎压对复合工况轮胎力学特性影响的研究还较为匮乏。为描述不同胎压下的复合工况轮胎力学特性，建立考虑胎压影响的UniTire复合工况轮胎模型。通过试验观察和理论模型分析相结合的方法，揭示胎压和载荷耦合效应对轮胎侧纵向滑移刚度等关键力学特性的影响规律及其机理。进一步通过分析胎压载荷耦合影响下复合工况轮胎接地印迹内总切力方向的变化规律，完善总切力方向因子表达式，建立考虑胎压载荷耦合影响的复合工况轮胎模型。利用轮胎试验数据对模型进行广泛验证，结果表明所建立的UniTire复合工况模型能精确地描述大范围胎压和载荷变化下的轮胎力学特性。

The change of inflation pressure and vertical load will significantly affect the mechanical characteristics of tires, and then affect the handling stability of vehicles. However, the current research on the effect of inflation pressure on the mechanical characteristics of tires under combined slip conditions is still insufficient. In order to describe the mechanical characteristics of tires under different inflation pressures, the UniTire combined slip model considering inflation pressure effects is established. By means of experimental observation and theoretical model analysis, the influence and mechanism of the coupling effect of inflation pressure and vertical load on the key mechanical characteristics such as cornering and longitudinal slip stiffness of tires are revealed; Furthermore, by analyzing the change of the resultant shear force direction in the contact patch under combined slip conditions and the influence of the coupling effect of inflation pressure and vertical load, the expressions of the direction factor are improved, and a combined slip tire model is established considering the influence of the coupling effect of inflation pressure and vertical load. The model is extensively validated using tire test data, and the results show that the UniTire combined slip model can accurately describe the tire mechanical characteristics under a wide range of inflation pressure and vertical load changes.

机电传动系统在新能源车辆、高速列车等领域广泛应用，随着系统向高速、重载和轻量化方向的发展，其动力学特性愈发复杂。基于集中-分布参数模型，分析渐开线直齿轮啮合激励和轴的振动激励共同作用情况下的动力学响应及影响规律。基于有限单元法和集中参数法，结合轴模型和齿轮啮合模型建立集中-分布参数动力学模型，采用广义-a数值方法进行求解，分析轴振动和齿轮偏心量对系统动力学响应的影响。研究发现：集中-分布参数模型能够预测出轴在齿轮安装位置和轴与轴承一侧连接位置两点处承受较大的von Mises交变应力；轴的变形可诱发系统产生由轴沿Y轴、Z轴方向的弯曲振动频率及其与啮频的组合频率等新频率；仿真得到轴沿Y轴和Z轴方向的弯曲振动频率分别为1 125 Hz和874 Hz，两个频率的大小与幅值仅由轴的固有特性决定；随着齿轮偏心量增大，时变齿侧间隙、动态啮合力、轴上随时间周期性变化的von Mises交变应力和齿轮中心涡动轨迹均明显增大，且最大动态啮合力与齿轮偏心量呈线性递增关系，递增比率为1.77×10⁹. 该研究成果可为提升机电传动系统的动力学性能提供重要的理论参考依据。

针对野外环境中自动驾驶履带车辆轨迹跟踪控制问题，考虑建模误差、参数不确定性及外界随机强干扰，以强鲁棒性及精确跟踪为目标，提出一种基于误差符号鲁棒积分的自动驾驶履带车辆鲁棒自适应轨迹跟踪控制方法。基于拉格朗日动力学方程建立自动驾驶履带车辆的运动学与动力学耦合模型；采用自适应控制方法实现对模型的精确前馈补偿，抵消模型非线性的影响；通过误差符号鲁棒积分有效抑制外界干扰及不确定性的影响；利用Lyapunov稳定性理论证明了闭环系统的全局渐进稳定性与收敛性。对仿真结果进行了实车实验一致性验证。仿真和实验结果证明：该方法在存在建模误差、参数不确定性、外界干扰条件下，在实现自动驾驶履带车辆高精度轨迹跟踪控制的同时，具有较强的自适应和鲁棒性。

电驱动履带车辆具有良好的运动可控性，同时可借助电气制动缓解传统履带车辆制动系统负荷重、寿命短的问题，是履带车辆实现无人驾驶的理想驱动方式。通过对某电驱动履带车辆制动系统的无人化设计研究，提出了一种机电联合制动系统线控化的完整技术方案。该方案采用一种改进的三段式机械-电气制动结合方式，并在保证既定制动性能前提下按照最大化制动能量回收的原则，给出了相应的机械-电气制动力分配策略。按照该方案进行平台搭建后，进行了制动性能实车试验，验证了该系统具有良好的制动性能和工作稳定性，可在充分满足国家军用标准对军用履带车辆制动性能要求的同时，保证整体效率在25%左右的动能转化效率。

The electric drive tracked vehicle has good motion controllability, and the problems of heavy load and short life of brake system for the traditional tracked vehicle can be relieved by means of electric brake, which is an ideal driving method for realizing the unmanned driving of tracked vehicle. Through the design of braking system for an electric drive unmanned tracked vehicle, a complete technical solution for the electro-mechanical combined braking system is proposed, which is totally controlled by wire. An improved three-stage mechanical-electric brake combination method is adopted, and according to the principle of maximizing braking energy recovery under the premise of ensuring the custom dynamic performance, a corresponding mechanical-electrical braking force distribution strategy is given. After the platform was built according to the scheme, the brake performance test was carried out. The test results show that the system has good braking performance and stability, and can fully meet the requirements of the national military standard for the braking performance of military tracked vehicles while ensuring the kinetic energy conversion efficiency being around 25%. Key

The energy management strategy based on reinforcement learning encounters the problem of “dimension disaster”when dealing with high-dimensional problems because of the discretization of state and control variables. For this problem, a new energy management algorithm based on deep reinforcement learning with normalized advantage function is proposed, where two deep neural networks with normalized advantage function are used to realize the continuous control of energy and eliminate the discretization of state and control variables. Based on the modeling of powertrain of a series hybrid tracked vehicle, the framework of the proposed deep reinforcement learning algorithm was built and the parameter update process was completed for the series hybrid tracked vehicle. The simulated results show that the proposed algorithm can output more refined control quantity and less output fluctuation. Compared with the deep Q-learning algorithm, the proposed algorithm improves the fuel economy of series hybrid tracked vehicle by 3.96%. In addition, the adaptability of the proposed algorithm and the optimized effect in real-time control environment are verified by the hardware-in-the-loop simulation.

类人驾驶系统是通过学习人类驾驶员知识与经验来提升无人驾驶系统适用性与接受度的重要技术途径。为解决驾驶员轨迹和操控层面经验的表述问题，以采集得到的大量真实驾驶数据为依托，提出一种基于轨迹基元与操控基元的分层式驾驶员经验表述模型。轨迹基元以动态运动基元算法进行表征，并由概率提取算法完成基元从无标签连续轨迹数据中的分割提取。操控基元在轨迹基元的提取分类结果上，利用高斯混合模型完成基元的训练，并利用高斯回归算法完成转向操控序列的预测。结果表明，概率提取算法既利用到了表征与提取之间的相互关联关系，又借助于初始分割点的合理设置，提升了算法的效率并使得提取得到的运动基元符合特定的驾驶假设。此外，所提出的运动基元既能以较高精度完成对驾驶员轨迹和操控层面数据的表征，又具备良好的泛化能力以应对运动基元再生成时在期望位置和时间尺度上的变化需求。最终构建了描述全工况驾驶行为的运动基元库，并大幅提升了运动基元应对不同行车环境时的适用性。

The human-like driving system is an essential technical way to improve the applicability and acceptance of an unmanned driving system by learning the knowledge and experience of human drivers. In order to solve the driving skills representation problem at trajectory and control level, by utilizing a large amount of collected real driving data, a hierarchical driver model based on trajectory primitives and operation primitives is proposed. The trajectory primitives are represented by the dynamic movement primitive, and the probabilistic extraction algorithm is applied to extract primitives from the unlabeled continuous trajectory data. The operation primitives use the Gaussian mixture model to complete the training process based on the extraction and classification results of the trajectory primitives. The Gaussian mixture regression(GMR) algorithm is applied to predict the steering angle. The results show that the probabilistic extraction algorithm not only utilizes the correlation between representation and extraction but also uses the reasonable setting of the initial segmentation point, which improves the efficiency of the algorithm and makes the extracted motion primitives conform to specific driving assumptions. The proposed motion primitives can not only represent the driver's driving data with high precision but also have good generalization ability to deal with the desired position and time duration change when the motion primitives are regenerated. Finally, the motion primitive library describing the driving behavior under all conditions is established, and the applicability of the motion primitives to different driving situations is significantly improved.

A norm version of the RMSProp algorithm with penalty (termed RMSPropW) is introduced into the deep learning framework and its convergence is addressed both analytically and numerically. For rigour, we consider the general nonconvex setting and prove the boundedness and convergence of the RMSPropW method in both deterministic and stochastic cases. This equips us with strict upper bounds on both the moving average squared norm of the gradient and the norm of weight parameters throughout the learning process, owing to the penalty term within the proposed cost function. In the deterministic (batch) case, the boundedness of the moving average squared norm of the gradient is employed to prove that the gradient sequence converges to zero when using a fixed step size, while with diminishing stepsizes, the minimum of the gradient sequence converges to zero. In the stochastic case, due to the boundedness of the weight evolution sequence, it is further shown that the weight sequence converges to a stationary point with probability 1. Finally, illustrative simulations are provided to support the theoretical analysis, including a comparison with the standard RMSProp on MNIST, CIFAR-10, and IMDB datasets.Copyright © 2021 Elsevier Ltd. All rights reserved.