Integrating self-health awareness in autonomous systems
Introduction
Autonomous unmanned vehicles will require intelligent vehicle management systems capable of adapting to unanticipated operating conditions. Autonomous system approaches have traditionally struggled with the representation of the external environment in which the system operates. An accurate representation of the internal state – including the health of critical subsystems – can be equally challenging. By integrating knowledge of the health and capability of critical subsystems into the intelligent control system for an autonomous vehicle, the system can react to both external and internal changes. The system must be capable of using internal and external situational awareness to determine whether current system capability exceeds, meets, or fails to meet current demands. If capability does not meet or exceed demands, the system could become damaged or incapacitated as a result of executing the current mission plan. In fully autonomous systems, a human operator may not be able to intervene or rescue the system; hence levels of damage or incapacitation that would be considered repairable in a manned system may be fatal to the mission of an autonomous system.
Internal self-situational awareness implies that the autonomous control system not only knows the health and performance characteristics of critical components or subsystems, but that the system can use that information to determine whether current capability meets or exceeds current demands. Assessing whether current capability exceeds mission demands requires two pieces of information:
- •
Expected demands on the critical components and subsystems.
- •
Current subsystem or component health and performance.
While many researchers have focused on developing tools and techniques for diagnosing problems in mechanical and electrical systems, determining the loads and demands on critical components and subsystems based on the planned mission or operations is itself a nontrivial problem.
The ability to respond to unanticipated changes in system health and performance is important not only for standalone autonomous systems, but also for teams of collaborating autonomous systems. By sharing information on the health and capability of individual team members, a collaborating team of autonomous systems can optimize usage of communal resources and adjust roles and responsibilities of individual members to insure mission success.
Information on the health and capability of critical subsystems can be integrated into an autonomous control system in a number of ways, each requiring different amounts of additional sensor or communication bandwidth and requiring additional processing resources. At one end of the spectrum, data from sensors monitoring the health and performance of critical subsystems and components can be provided directly to the autonomous control system for incorporation into the control decision making process. At the other end of the spectrum, a parallel, autonomous, health monitoring system can perform all assessments of subsystem and component health and provide high-level advisory inputs to the autonomous control system. The representation of knowledge about system operation and capability must be consistent between the health monitoring system and the autonomous control system. Representation of subsystem and component health must be in a form that is useable to the autonomous control system without significantly adding to the computational or communication bandwidth requirements of the system.
This paper describes the later approach to integrating health monitoring and control and explores a method for integration of the health information into a representative autonomous intelligent control system. The integration of the health information into the autonomous control system is demonstrated using a behavior based autonomous intelligent control system architecture used by the Applied Research Laboratory in autonomous underwater vehicle applications. The application of integrated health monitoring and autonomous control is demonstrated in a simulation of two autonomous underwater vehicles executing a joint search mission. The application of integrated health monitoring and control is also described for an autonomous ground vehicle platform.
Section snippets
System health monitoring for self-awareness
Automated system health monitoring involves the application of sensors, analysis, data fusion and automated reasoning to estimate the health and track the degradation of a system. A top-level view of causes and effects in system health monitoring is shown in Fig. 1. Demands and loads propagate from mission requirements at the platform level down to the material level; the failure and degradation of components and subsystems start at the material level and propagate up to affect operational
Autonomous intelligent control architecture
A behavior-based, autonomous intelligent control architecture is used to integrate internal and external self-situational awareness and execute the desired mission. In fact, the same architecture is used within the health monitoring subsystem to provide the internal self-situational awareness and to provide external situational awareness. The use of common autonomous system architectures eases the task of integrating system monitoring with control to provide autonomous vehicle management.
The
Integrating health monitoring into autonomous systems
Designers have a number of choices when integrating health information into an autonomous system. Health information can be integrated in the form of low-level sensor data by providing data from sensors mounted on critical components or subsystems directly into the autonomous control system. At the other extreme, health information can be integrated as high level information on the capability of critical components or subsystems. The latter approach reduces the bandwidth required for the
Multi-vehicle collaboration
A key issue in multi-vehicle systems is the level of communication required to control the action of the collective. Although many possible configurations exist, there are three architectures that are worth considering:
- •
No direct communication.
- •
All communication routed through a supervisor.
- •
Fully integrated communication among peers.
An additional consideration that determines how much information needs to be shared is whether the collective of vehicles is homogeneous or heterogeneous. The
Conclusions
In order to achieve high levels of safety and mitigate mission risk, unmanned systems will require self-situational awareness and autonomous decision and response capability to respond to unplanned events, system degradation and failure. The Applied Research Laboratory has developed and demonstrated both system health monitoring and autonomous intelligent control systems for a variety of platforms and applications. The integration of these technologies provides an enabling capability for
Acknowledgements
This paper is based upon work supported by NASA under the Engineering for Complex Systems Program—Autonomous Propulsion Systems Technology Subprogram (Grant No. NAG3-2778).
Karl Reichard has more than 15 years of experience in the development of advanced sensors, measurement systems, and signal processing algorithms. An Assistant Professor of Acoustics at Penn State and Head of the Condition-Based Maintenance Department at the University's Applied Research Laboratory, Dr. Reichard leads advanced research and development efforts in intelligent health monitoring and control systems, embedded systems, intelligent acoustic and vibration sensors, and signal processing
References (4)
- et al.
Expanding the foundation for prognostic health management in complex mechanical systems
- et al.
Self-awareness, monitoring and diagnosis for autonomous vehicle operations
Cited by (17)
Research and application on rapid recognition of CNC machine tools' running state
2020, IOP Conference Series: Materials Science and EngineeringDesigning transparent and autonomous intelligent vision systems
2019, ICAART 2019 - Proceedings of the 11th International Conference on Agents and Artificial IntelligenceHealth evaluation and coordination control for manufacturing system
2016, Jiangsu Daxue Xuebao (Ziran Kexue Ban)/Journal of Jiangsu University (Natural Science Edition)Towards functional safety in drive-by-wire vehicles
2015, Towards Functional Safety in Drive-by-Wire Vehicles
Karl Reichard has more than 15 years of experience in the development of advanced sensors, measurement systems, and signal processing algorithms. An Assistant Professor of Acoustics at Penn State and Head of the Condition-Based Maintenance Department at the University's Applied Research Laboratory, Dr. Reichard leads advanced research and development efforts in intelligent health monitoring and control systems, embedded systems, intelligent acoustic and vibration sensors, and signal processing and classification algorithms for active noise and vibration control, manufacturing machinery monitoring, and surveillance systems. Prior to joining Penn State ARL in 1991, he was employed by the U.S. Army Aberdeen Proving Grounds and Virginia Polytechnic Institute and State University, his alma mater. Dr. Reichard has published more than 25 papers in refereed journals, conference publications, and technical reports.