基于车辆动力学模型的换道轨迹规划研究

针对自动驾驶车辆换道轨迹规划时的操纵稳定性问题，基于CarSim/Simulink仿真平台建立了车辆动力学模型，构建了轨迹规划系统框架，通过轨迹信息后处理并提出了目标函数设计，进行了横向控制序列采样以保证车辆的稳定与极限性能，完成了算法对轨迹的综合评价选优。随后开展了仿真试验，对比分析了轨迹跟踪控制系统下的实际轨迹、最优规划方法所规划的换道轨迹。仿真结果表明，该轨迹规划系统框架及算法模型能有效提高车辆的操纵稳定性，可实现冰雪路面等极端工况下自动驾驶车辆换道轨迹规划。     

基于稀疏卷积神经网络的车载激光雷达点云语义分割方法

对车载激光雷达扫描得到的点云进行语义分割是保证行车安全、加强驾驶员对周边环境理解的重要手段之一。因为内存限制和大规模点云场景更加稀疏的特点,将传统神经网络的方法直接沿用到车载激光雷达扫描得到的点云场景中的效果不佳。本文中针对大规模点云的稀疏性,利用稀疏卷积神经网络对体素化点云进行特征提取。考虑到逐点处理分支抑制点云数据的密度不一致性导致的信息损失,另外设计了3D-CA和3D-SA模块,使稀疏卷积神经网络更好地提取特征。实验结果表明,与传统卷积神经网络的方法和将点云投影到平面的方法相比,使用稀疏卷积神经网络对大规模点云进行语义分割,可将平均交并比提升4.1%和3.4%,证明了该方法的有效性。     

Anti-disturbance optimal coverage control of ASVs北大核心CSCD

[目的]针对固定的海洋区域,研究欠驱动水面无人船(ASV)集群的抗干扰最优覆盖控制问题。[方法]首先,在运动学层级,基于邻居ASV位置和环境密度信息进行Voronoi分配,设计欠驱动ASV运动学制导律,引导ASV运动到最优目标点;然后,考虑复杂海洋环境下ASV模型的不确定性以及风浪流引起的外部扰动,在动力学层级设计一种固定时间扩张状态观测器(FTESO),并基于FTESO设计固定时间的动力学控制律。[结果]稳定性分析结果表明了ASV集群的抗干扰最优覆盖控制闭环系统的误差是有界的,仿真结果验证了该方法的有效性。[结论]通过采用该抗干扰最优覆盖控制器,ASV集群可以在任意的初始位置实现对目标区域的最优覆盖。     

Seabed collision emergency decision-making of AUV based on safety domain model北大核心CSCD

［Objectives］To ensure safety and prevent seabed collisions in complex unknown underwater environments, this study proposes a seabed safety domain model and tiered emergency response strategies. ［Methods］A vertical motion simulation model is established and verified by surpassing the test results, then used to calculate the active and passive safety domain distance of an autonomous underwater vehicle (AUV), thereby establishing a seabed safety domain model. An AUV emergency control system and emergency strategies are then built on the basis of the dynamic safety domain model. The trim and distance from the seabed of the AUV are used to calculate the current and future risk factors. Based on the weighted sum, the comprehensive risk factor is employed to provide the AUV with emergency response strategies.［Results］Lake tests with the AUV sailing at a fixed depth and height show a strong dependency of the comprehensive risk coefficient on seabed height when it is close to the boundary of the AUV's active safety domain. In the opposite case, there is a weak dependency of the comprehensive risk coefficient on seabed height. The results show that the proposed AUV emergency control system can reduce emergency false alarms caused by frequently changing riverbed heights and sailing altitudes close to the seabed. In such cases, reasonable emergency strategies can be realized under complex rough terrain.［Conclusions］The AUV seabed safety domain model and tiered emergency response strategies based on vertical motion equations proposed herein can be applied to evaluate seabed collision risk in various cases. Finally, this paper provides emergency response strategies to avoid seabed collision accidents, which can enhance the safety of AUV navigation. © 2023 Chinese Journal of Ship Research. All rights reserved.