Drift Counteraction Control: Theory and Applications

漂移抵消控制：理论与应用

• 批准号: 1404814
• 负责人: Ilya Kolmanovsky
• 金额: $ 20万
• 依托单位: Regents of the University of Michigan - Ann Arbor
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-08-15 至 2018-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1404814&HistoricalAwards=false
• 关键词: Drift; Counteraction; Control; Theory; Applications

Drift Counteraction Control: Theory and ApplicationsIn many existing and emerging engineering systems there is an inherent tendency of process variables to drift. In some systems this drift is caused by large, persistent disturbances due to interactions with the external environment. In other systems, similar drift is caused by finite resources (fuel, energy, component life etc.) being continuously depleted. The operating objectives for such systems with drift are reflected in a set of constraints that must be satisfied for as long as possible by countering the drift, rather than in terms of usual set-point command tracking requirements. This research project advances theory and methods for designing control algorithms that perform drift counteraction thereby establishing a foundation for addressing practical drift counteraction applications, and for incorporating drift counteraction technology into industrial products. These advances will represent contributions to several areas of control theory including stochastic control, set-theoretic control, game-theoretic control, constrained control, and nonlinear control. The developments in theory will proceed in close synergy with the demonstration of their benefits in several engineering applications to automotive and aerospace systems and dynamic motion simulators. The research results will be integrated into university courses for graduate students, short courses for industrial and academic audiences, and into the computational software package implementing drift counteraction control design techniques. The drift counteraction control is based primarily on constraints and does not use conventional set-points, differently from most of the traditional control theory. The development of effective techniques for drift counteraction will be based on treating external disturbances causing drift and constraint violation first in the stochastic and then in the deterministic setting. In the stochastic setting, advances in Stochastic Drift Counteraction Optimal Control (SDCOC) that maximizes the expected cost before the imposed constraints are violated will be pursued. Stability and boundness conditions for closed-loop trajectories of systems operating under SDCOC control laws when disturbances settle to constant values or vary within a subset of the full range will be derived. Sub-optimality and convergence properties of receding horizon and reinforcement learning variants of SDCOC will be studied. Advances in SDCOC computational procedures will be made, firstly, by exploiting decomposition of the system dynamics into slow and fast subsystems and, secondly, based on Markov Chain models that use Fuzzy Encoding (MCFE). By combining SDCOC and MCFE, evidence based control framework for drift counteraction in uncertain systems will be defined and complemented with estimation algorithms. In the deterministic setting, set-theoretic, game-theoretic and disturbance estimation/cancellation based approaches to drift counteraction problems will be pursued. The advances will focus on formulation and derivation of algorithms, the improvements in their off-board and on-board computations, and characterization of closed-loop properties (constraint violation time, ability to prevent constraint violation over an infinite time interval, trajectory bounds, etc.). Synergistically with the theoretical advances, several practical applications of drift counteraction control will be considered. These applications will include increasing comfort and fuel efficiency of adaptive cruise control systems for conventional and hybrid passenger vehicles, extending range and life of electric and hybrid electric vehicles and their batteries, improving the ability to replicate motions in small scale dynamic motion simulators, and improving the capability and efficiency of spacecraft attitude control using momentum exchange devices. For each application, its requirements will be reflected in a drift counteraction problem formulation, appropriate models will be established, drift counteraction control algorithms will be designed, and performance benefits will be quantified. To ensure practical relevance of the results and facilitate their experimental validation in the vehicle adaptive cruise control case, interactions and collaborations with industry partners will be leveraged.

漂移对抗控制：理论与应用在许多现有的和新兴的工程系统中，过程变量具有漂移的固有趋势。在某些系统中，这种漂移是由与外部环境相互作用产生的大而持久的干扰引起的。在其他系统中，类似的漂移是由有限的资源（燃料、能源、组件寿命等）不断耗尽引起的。这种具有漂移的系统的操作目标反映在一组限制条件中，这些限制条件必须尽可能长时间地通过对抗漂移来满足，而不是按照通常的设定点命令跟踪要求。本研究项目推进了设计执行漂移抵消的控制算法的理论和方法，从而为解决实际的漂移抵消应用以及将漂移抵消技术纳入工业产品奠定了基础。这些进展将对控制理论的几个领域做出贡献，包括随机控制、集合理论控制、博弈论控制、约束控制和非线性控制。理论的发展将与它们在汽车和航空航天系统以及动态运动模拟器的几个工程应用中的益处的演示密切协同进行。研究成果将整合到研究生的大学课程，工业和学术受众的短期课程，以及实现漂移抵消控制设计技术的计算软件包中。与大多数传统控制理论不同，漂移抵消控制主要基于约束，不使用传统的设定点。有效的漂移对抗技术的发展将基于先在随机环境中处理引起漂移和违反约束的外部干扰，然后再在确定性环境中处理。在随机环境下，我们将继续研究随机漂移对抗最优控制（SDCOC）的研究进展，以期在违反所施加的约束之前使期望成本最大化。在SDCOC控制律下，当扰动稳定在恒定值或在全范围的一个子集内变化时，将推导出系统闭环轨迹的稳定性和有界条件。研究了SDCOC的后退视界和强化学习变体的次优性和收敛性。SDCOC的计算过程将取得进展，首先，通过将系统动力学分解为慢速和快速子系统，其次，基于使用模糊编码（MCFE）的马尔可夫链模型。通过结合SDCOC和MCFE，定义了不确定系统漂移抵消的基于证据的控制框架，并辅以估计算法。在确定性的设置，集理论，博弈论和干扰估计/消除为基础的方法漂移抵消问题将被追求。这些进展将集中在算法的制定和推导，其机载和机载计算的改进，以及闭环特性的表征（约束违反时间，在无限时间间隔内防止约束违反的能力，轨迹边界等）。与理论进展协同，漂移抵消控制的几个实际应用将被考虑。这些应用将包括提高传统和混合动力乘用车自适应巡航控制系统的舒适性和燃油效率，延长电动和混合动力汽车及其电池的续航里程和寿命，提高小型动态运动模拟器复制运动的能力，以及使用动量交换装置提高航天器姿态控制的能力和效率。对于每个应用，其需求将反映在漂移抵消问题的表述中，建立适当的模型，设计漂移抵消控制算法，并量化性能效益。为了确保结果的实际相关性，并促进其在车辆自适应巡航控制案例中的实验验证，将利用与行业合作伙伴的互动和合作。