GOALI: Modeling, Evaluation, and Control of Tire Blowout for Automated and Partially Automated Vehicles

该项目为建立有效的系统来模拟、预测和减轻轮胎爆胎对汽车动力学的影响奠定了研究基础。该项目的目标是大幅提高道路安全，特别是在驾驶辅助系统或全自动驾驶在个人车辆中越来越普遍的情况下。讨论了三个具体的研究重点。第一个研究重点是对爆裂事件本身进行深度建模——本质上是对轮胎迅速失去完整性时发生的变化进行“超级慢动作”观察。该模型嵌入到车辆仿真软件中，用于预测各种车辆在不同环境条件下的动态响应。由于这种高保真度的建模方法将井喷的短暂持续时间划分为许多微小的时间间隔，因此所得的计算机软件速度太慢，无法在实际事件中使用。因此，第二个研究重点是创建一个轮胎爆胎模型，该模型可以由车载计算机运行，以帮助控制汽车。“脉冲”控制模型将井喷事件压缩为车辆状态的单个跳跃变化。车载计算机使用脉冲模型来解读传感器读数，确定已经发生的井喷类型，并预测汽车随后的运动。最后，车载计算机使用该模型控制汽车，直到它可以安全停车。这种控制器适用于自动驾驶汽车，但在不久的将来，更有可能看到部分自动驾驶的部署，其中驾驶员辅助功能与人类操作员共享一定程度的控制。对于部分自主情况，车载控制器必须能够预测和适应操作员的输入。因此，该项目的第三个研究重点是创建一个与车辆控制系统集成的人类行为模型，这将实现有效的共享控制，并确保在各种驾驶员技能和情绪状态下安全运行。这一项目的成功成果每年可防止数万起事故和数百人死亡。与通用汽车公司的研究伙伴关系将促进结果向商业实践的转变，通用汽车公司将提供主题专业知识，并在通用汽车工厂进行全面测试，验证比例模型实验室实验。三个任务中的第一个将推进研究团队的初步努力，以实验验证，高保真的轮胎爆胎模型。该模型对爆胎过程中的受力进行了两阶段分析，第一阶段考虑轮胎的初始塌陷，第二阶段考虑轮辋接触地面后的车轮行为。这个高保真模型将通过轮胎压力、轮胎尺寸、转弯刚度和车辆几何形状等参数化，并将包括对车辆行为对这些参数的敏感性分析。轮胎爆胎动力学与前后车辆动力学之间的大时间尺度分离使得该模型对于实时使用来说过于笨拙。因此，第二项任务是推导并验证爆胎前后车辆动力学的脉冲模型。也就是说，爆胎事件本身的动力学被忽略了，它的影响被集中成施加在受损轮胎上的瞬时脉冲力，加上车辆参数的相关阶跃变化。对于两种类型的估计器，将获得模型的版本——已知爆胎身份的情况，以及必须推断的情况。考虑完全自动驾驶和部分自动驾驶汽车都很有趣。对于后者，人类驾驶员与自动控制器共享控制。在共享控制下，自动控制器应该能够解释和预测人类驾驶员的输入，以便提供适当的稳定响应。因此，第三个研究任务是创建一个由车辆控制器实现的人类行为模型。所设想的模型使用带有时间延迟的三阶传递函数来表示人类对车道偏离误差的动态响应。随后，该项目将开发控制器，以确保自动驾驶和部分自动驾驶车辆爆胎后的车道维护安全。这项任务的创新包括推导结合脉冲和连续控制和干扰的动态模型。该项目还将开发脉冲可观测性工具，以描述可以从传感器测量推断出井喷特征的条件。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This Grant Opportunities for Academic Liaison with Industry (GOALI) project advances a research foundation for effective systems to model, predict, and mitigate the effects of tire blowout on automobile dynamics. The goal of this project is to substantially improve road safety, especially as driver-assist systems or full autonomy become more widespread in personal vehicles. Three specific research thrusts are addressed. The first research thrust is an in-depth model of the blowout event itself -- essentially a "super slow-motion" look at the changes that occur as the tire rapidly loses its integrity. This model is embedded in a vehicle simulation software, to predict the dynamic response of various vehicles, under different ambient conditions. Because this high-fidelity modeling approach splits the brief duration of the blowout into many tiny time intervals, the resulting computer software is too slow to be used during an actual event. Therefore, the second research thrust is to create a model of tire blowout that can be run by an on-board computer to help keep the car under control. An "impulsive" control model condenses the blowout event into a single jump change in the vehicle status. An on-board computer uses the impulsive model to interpret sensor readings, determine the type of blowout that has occurred, and predict the subsequent motion of the car. Finally, the on-board computer uses the model to keep the car under control until it can be brought to a safe stop. This controller is appropriate for self-driving vehicles, however the near future is more likely to see deployment of partial autonomy, in which a driver-assist feature shares some degree of control with a human operator. For partially autonomous situations the on-board controller must be able to predict and accommodate operator inputs. Therefore, the third research thrust of the project is the creation of a human-behavior model to be integrated with the vehicle control system, that will enable effective shared control and ensure safe operation over a wide range of driver skills and emotional states. Each year tens of thousands of accidents and hundreds of deaths could be prevented by a successful outcome to this project. The transition of the results to commercial practice will be facilitated by a research partnership with General Motors, who will provide subject-matter expertise, and validate scale-model laboratory experiments with full-scale testing at GM facilities.The first of three tasks will advance the research team's preliminary efforts towards an experimentally validated, high-fidelity model of tire blowout. The model features a two-phase analysis of forces during blowout, with the first phase accounting for the initial collapse of the tire, and the second phase accounting for the wheel behavior after the rim contacts the ground. This high-fidelity model will be parametrized by quantities such as tire pressure, tire size, cornering stiffness, and vehicle geometry, and will include an analysis of the sensitivity of vehicle behavior to those parameters. The large time-scale separation between the dynamics of the tire blowout and the preceding and subsequent vehicle dynamics makes this model too unwieldy for real-time use. Therefore, the second task is to derive and validate an impulsive model of the vehicle dynamics before and after tire blowout. That is, the dynamics of the blowout event itself are ignored, and its effects are gathered into an instantaneous impulsive force applied at the stricken tire, plus an associated step change in the vehicle parameters. Versions of the model will be obtained for two types of estimators -- the case where the identity of the blown-out tire is known, and the case for which it must be inferred. It is of interest to consider both fully autonomous and partially autonomous vehicles. For the latter, a human driver shares control with an automatic controller. Under shared control, the automatic controller should be able to interpret and predict the human driver input, in order to provide an appropriate stabilizing response. Thus the third research task is to create a model of human behavior, implementable by the vehicle controller. The envisioned model represents the dynamics of human response to lane departure error using a third-order transfer function with time delay. Subsequently, the project will derive controllers to ensure safe lane maintenance after tire blowout for both automated and partially automated vehicles. Innovations in this task include the derivation of dynamic models combining impulsive and continuous controls and disturbances. The project will also develop tools of impulsive observability to describe conditions under which the blowout characteristics may be inferred from sensor measurements.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.