Cooperative Fault Diagnosis and Control of a Network of Unmanned Systems

最近在控制、计算和通信方面的技术进步引起了人们对开发新一代高度互联、分布式、传感器丰富的自主系统的浓厚兴趣，这些系统适用于各种复杂环境，例如用于情报、监视和侦察（ISR）的无人驾驶车辆(如无人驾驶飞机、无人驾驶水下和无人驾驶地面车辆（uav / uuv / ugv）、卫星和太空飞行器（SVs）的合作团队。自然灾害和环境监测，科学的地球和外层空间观测和发现，等等。由于执行器和传感器的异常和故障等不良影响，必须监测无人系统（UMS）合作团队的健康状况，以便这些所谓的“代理”能够诊断故障，响应和重新配置控制律（控制恢复和可重构性），达成共识，并针对不可预见和不可预测的情况提出行动方案。因此，有必要协作调整控制律以使代理从异常、故障和失败的影响中恢复过来。具体而言，需要开发自主智能的分布式故障诊断（FD）和故障重构（FR）控制系统来管理UMS/SVs的网络化团队。特别是，如果某个agent发生故障，而团队又不具备自主容错控制能力，那么团队的稳定性、共识和形成就可能无法维持，从而导致整个团队的不稳定。拟议研究的主要目的是推广和扩展FD&R技术和方法，从单一的无人系统到合作的UMS/SVs团队，以完成给定的任务。故障诊断和恢复（FD&R）技术已经研究并可用于单个移动机器人系统，然而，FD&R技术并不常见，并且在UMS/SVs团队中得到了广泛的研究和可用，这是本提案的主要区别和独特性。提出的研究目标是为自主车载系统提供增强的能力，以确定最有效的分配、配置、策略、算法和架构，以进行交互式UMS/ sv网络的协同故障诊断和控制。设计和开发一个自组织和自适应的UMS/SVs网络，自主地展示新的集体行为，提供了一些目前使用单个代理无法获得或无法实现的优势和机会。在这个项目中设想和提出的基本挑战和问题是开发新颖的、正式的、严格的分布式和半分散的方法，用于管理和确定合作的UMS/SVs团队之间的数据、知识和行动的交换。这些目标是通过利用受通信网络和计算资源约束的控制方案来实现协同故障诊断、协同恢复和重新配置控制目标，以保证和维持任务要求。提出了利用决策理论、非线性鲁棒自适应、协同与协作分布式故障诊断与控制技术的创新研究方法。这项研究将为加拿大政府实验室和工业，特别是航空航天工业的重要实际问题提供严格的解决方案，并将积累人力资本和发展可持续的研究能力。

Recent technological advances in control, computing, and communications have generated intense interest in development of a new generation of highly interconnected, distributed, sensor rich autonomous systems for applications in a variety of complex settings such as cooperating team of unmanned vehicles (such as unmanned aerial, unmanned underwater, and unmanned ground vehicles (UAVs/UUVs/UGVs), satellites, and space vehicles (SVs)) for intelligence, surveillance, and reconnaissance (ISR), natural disaster and environmental monitoring, scientific Earth and outer space observations and discovery, to name a few. Due to undesirable effects such as anomalies and faults in actuators and sensors, the health of cooperating team of unmanned systems (UMS) must be monitored so that these so-called “agents” can diagnose faults, respond and reconfigure control laws (control recovery and reconfigurability), reach a consensus, and propose a course of action to unforeseen and unpredictable circumstances. Consequently, it will be necessary to cooperatively adjust control laws to recover the agents from the effects of anomalies, faults, and failures. Specifically, what is required is to develop autonomous and intelligent distributed fault diagnosis (FD) and fault reconfiguration (FR) control systems to manage the networked team of UMS/SVs. In particular, if any agent is subjected to a fault and the team is not equipped with autonomous fault tolerant control capabilities, the stability, consensus, and formation of the team may no longer be maintained and this could lead to instability of the entire team. The main thrust of the proposed research is to generalize and extend FD&R techniques and methodologies from a single unmanned system to a team of cooperating UMS/SVs tasked to accomplish a given mission. Fault diagnosis and recovery (FD&R) techniques have been studied and are available for individual mobile robotic systems, however, FD&R techniques are not common and extensively researched and available for a team of UMS/SVs, which is the main distinguishing aspect and uniqueness of this proposal. The objectives of the proposed research are to provide enhanced capabilities for autonomous on-board systems in determining the most efficient allocation, configuration, strategies, algorithms, and architectures for cooperative fault diagnosis and control of a network of interacting UMS/SVs. Design and development of a self-organizing and adaptive network of UMS/SVs that autonomously demonstrate novel collective behaviors offer several advantages and opportunities that are not currently available or possible by utilizing a single agent. The fundamental challenges and issues that are envisaged and proposed in this project are development of novel, formal, and rigorous distributed and semi-decentralized methodologies for management and determination of exchange of data, knowledge, and actions among the team of cooperating UMS/SVs. These goals are to be accomplished by utilizing control schemes that are constrained by the communication network and computational resources for cooperative fault diagnosis, cooperative recovery and reconfiguration control objectives for guaranteeing and maintaining the mission requirements. The proposed innovative research methodologies envisaged are to utilize decision theoretic, nonlinear robust and adaptive, collaborative and cooperative distributed fault diagnosis and control techniques. This research will provide rigorous solutions to practical problems of paramount importance that are of significant interest to Canadian government laboratories and industry in general and to the aerospace industry in particular, and will accumulate human capital and develop sustainable research capabilities.