Modeling and Motion Planning for Autonomous Skid-Steered Vehicles

这项研究致力于开发和实验验证滑动轮式车辆的三维运动动力学模型，这是一类大型和重要的户外车辆，由于其坚固的机械结构，经常被认为是全地形机动的首选。此外，本研究还将所开发的3D模型应用于运动规划。建模建立在严格的动力学分析基础上，但需要实验来确定关键参数，特别是剪切变形系数、摩擦系数和滚动阻力。研究还强调包括电机速度控制器，以获得对模型不确定性不那么敏感的闭环系统模型，从而为运动规划者提供更好的车辆性能预测。本研究中使用的主要运动规划方法称为基于采样的模型预测控制，它是模型预测控制的变体，它利用了采样方法的计算速度。滑动导向自主地面车辆(AGV)有望广泛用于非结构化、户外、起伏的环境中，执行搜索和救援、国土安全、消防、侦察和监视等任务。这些任务将大大受益于AGV的有效运动规划，这可以通过将SBMPC与本研究开发的车辆动力学模型相结合来完成。最终结果是时间最优的轨迹、高能效的运动规划、任务完成和加油的可靠能量预测、在存在非致命机械故障(例如爆胎)的情况下的运动规划，以及由于对车辆运动的更准确预测而较少频繁地重新规划。详细的建模、控制和规划程序预计将适用于RHex和Wheg机器人等日益增长和日益重要的防滑、独特的移动平台，这些平台由于更灵活的移动能力，例如攀登复杂障碍物的能力，预计将得到越来越多的使用。

This research focuses on developing and experimentally verifying dynamic models for 3D motion of skid-steered wheeled vehicles, a large and important class of outdoor vehicles, which are often preferred for all-terrain mobility due to a robust mechanical structure. In addition, the research will employ the developed 3D models in motion planning. The modeling is based on rigorous dynamic analysis, but requires experimentation to determine key parameters, particularly the shear deformation modulus, the coefficients of friction, and the rolling resistance. The research also emphasizes the inclusion of the motor speed controllers to obtain closed-loop models that are less sensitive to model uncertainty and can hence provide better predictions of vehicle performance for motion planners. The primary motion planning methodology used in this research is called Sampling Based Model Predictive Control and is a variant of model predictive control that takes advantages of the computational speed of sampling methods.Skid-steered autonomous ground vehicles (AGVs) are expected to be used extensively in unstructured, outdoor, undulating environments for tasks such as search and rescue, homeland security, fire fighting, reconnaissance, and surveillance. These tasks will greatly benefit from efficient motion planning for the AGVs, which can be accomplished by using SBMPC in conjunction with the dynamic models of the vehicles developed in this research. The end result is time-optimal trajectories, energy-efficient motion planning, reliable energy predictions for mission completion and refueling, motion planning in the presence of non-fatal mechanical failures (e.g., flat tires), and less frequent replanning due to more accurate predictions of vehicle motion. The detailed modeling, control and planning procedures are expected to be applicable to the growing and increasingly important class of skid-steered, unique mobility platforms such as RHex and Wheg robots, which are expected to find increasing use due to more flexible mobility, e.g., the ability to climb complex obstacles.