为解决虚拟仿真测试所需高覆盖场景的生成难题,以自动驾驶车辆事故高发的高速公路切入场景为研究对象,基于搭载多源传感器实车采集的自然驾驶数据,开发基于规则的切入场景自动提取算法;选取核心场景要素建立基于运动学特征的车辆切入轨...为解决虚拟仿真测试所需高覆盖场景的生成难题,以自动驾驶车辆事故高发的高速公路切入场景为研究对象,基于搭载多源传感器实车采集的自然驾驶数据,开发基于规则的切入场景自动提取算法;选取核心场景要素建立基于运动学特征的车辆切入轨迹模型,量化分析模型参数分布特征;基于高斯混合模型构建多维度逻辑场景参数联合概率密度函数,在此基础上提出基于哈密尔顿蒙特卡洛(Hamiltonian Monte Carlo,HMC)采样与Jensen-Shannon(JS)散度覆盖度表征的多维空间场景参数高覆盖生成方法。基于所提取的2422例车辆切入片段研究发现:①基于起始时刻主车速度Ve0、相对速度Vr0、车距Dx0、切入时长T、切入车辆横向加速度ay和纵向加速度ax六参数的横纵向运动学模型可有效表征切入车辆运动轨迹,平均拟合均方根误差为0.7 m;②八分量高斯混合模型对切入场景参数的联合概率密度分布拟合效果最佳;③JS散度随着场景采样数量的增加快速下降而后逐渐收敛至0.01,表明HMC方法可实现切入场景参数的快速采样与高覆盖生成;④本方法实现切入数据集片段信息全覆盖所需场景生成数量为2160个,相比于传统马尔科夫蒙特卡洛(Markov Chain Monte Carlo,MCMC)方法所需场景生成量缩小约73倍,测试效率显著提高,推荐用于高速公路全量切入场景库构建。提出的切入场景轨迹模型与高覆盖生成算法具有可解释性、高覆盖度、生成快捷的特点,将为自动驾驶虚拟仿真测试提供有力支撑。展开更多
The utilization of biomimicry of bacterial foraging strategy was considered to develop an adaptive control strategy for mobile robot, and a bacterial foraging approach was proposed for robot path planning. In the prop...The utilization of biomimicry of bacterial foraging strategy was considered to develop an adaptive control strategy for mobile robot, and a bacterial foraging approach was proposed for robot path planning. In the proposed model, robot that mimics the behavior of bacteria is able to determine an optimal collision-free path between a start and a target point in the environment surrounded by obstacles. In the simulation, two test scenarios of static environment with different number obstacles were adopted to evaluate the performance of the proposed method. Simulation results show that the robot which reflects the bacterial foraging behavior can adapt to complex environments in the planned trajectories with both satisfactory accuracy and stability.展开更多
文摘为解决虚拟仿真测试所需高覆盖场景的生成难题,以自动驾驶车辆事故高发的高速公路切入场景为研究对象,基于搭载多源传感器实车采集的自然驾驶数据,开发基于规则的切入场景自动提取算法;选取核心场景要素建立基于运动学特征的车辆切入轨迹模型,量化分析模型参数分布特征;基于高斯混合模型构建多维度逻辑场景参数联合概率密度函数,在此基础上提出基于哈密尔顿蒙特卡洛(Hamiltonian Monte Carlo,HMC)采样与Jensen-Shannon(JS)散度覆盖度表征的多维空间场景参数高覆盖生成方法。基于所提取的2422例车辆切入片段研究发现:①基于起始时刻主车速度Ve0、相对速度Vr0、车距Dx0、切入时长T、切入车辆横向加速度ay和纵向加速度ax六参数的横纵向运动学模型可有效表征切入车辆运动轨迹,平均拟合均方根误差为0.7 m;②八分量高斯混合模型对切入场景参数的联合概率密度分布拟合效果最佳;③JS散度随着场景采样数量的增加快速下降而后逐渐收敛至0.01,表明HMC方法可实现切入场景参数的快速采样与高覆盖生成;④本方法实现切入数据集片段信息全覆盖所需场景生成数量为2160个,相比于传统马尔科夫蒙特卡洛(Markov Chain Monte Carlo,MCMC)方法所需场景生成量缩小约73倍,测试效率显著提高,推荐用于高速公路全量切入场景库构建。提出的切入场景轨迹模型与高覆盖生成算法具有可解释性、高覆盖度、生成快捷的特点,将为自动驾驶虚拟仿真测试提供有力支撑。
基金Project(61173032)supported by the National Natural Science Foundation of ChinaProject(20090406)supported by the Tianjin Scientific and Technological Development Fund of Higher Education of China
文摘The utilization of biomimicry of bacterial foraging strategy was considered to develop an adaptive control strategy for mobile robot, and a bacterial foraging approach was proposed for robot path planning. In the proposed model, robot that mimics the behavior of bacteria is able to determine an optimal collision-free path between a start and a target point in the environment surrounded by obstacles. In the simulation, two test scenarios of static environment with different number obstacles were adopted to evaluate the performance of the proposed method. Simulation results show that the robot which reflects the bacterial foraging behavior can adapt to complex environments in the planned trajectories with both satisfactory accuracy and stability.
