Developing Knowledge and Understanding for Autonomous Systems for Analysis and Assessment Events and Campaigns

An effort to overcome the challenges involved in developing complete performance ontologies and test methodologies for evaluating the performance of autonomous systems.

There are many challenges in developing complete performance ontologies and test methodologies to define and evaluate the performance of autonomous systems. Chief among them is the dynamic environment in which the autonomous system is expected to operate. Change in the autonomous system’s environment is expected to affect system performance. Test methodologies have to include all aspects of this dynamic environment.

The ALFUS framework was developed by NIST to categorize autonomous systems in generic and metric-based terms. It is based on a three-axis model comprising the Mission Complexity, Environmental Complexity, and Autonomy Level axes.

Automation and autonomy offer significant military value in reducing danger to warfighters, in increasing the speed and accuracy of time-critical operations, and in reducing the supervisory burden of warfighters and control systems.

Autonomous systems can vary from simple adaptive automatic systems that are designed to operate in highly structured environments to fully self-governing systems designed to perform in highly dynamic and complex environments. Most Army autonomous systems are expected to range from automatic to semi-autonomous regimes.

Automated systems are systems that require little or no human involvement for performing well-defined tasks with predetermined responses. The system responses of automatic systems are generally rule-based and are designed to operate in well- structured environments with few parameters. Automated systems can be adaptive through the use of environmental sensors and rule-based adaptation. An example of an automatic system is a laundry machine that adapts to different laundry loads.

Autonomous systems are characterized as self-governing toward accomplishing their mission. System self-governance reduces the burden of control on the operator, allowing him or her to perform other tasks. Autonomous systems can be broadly classified by their level of autonomy:

  • Semi-Autonomous Systems perform limited control activities to react to changes in the environment. Automated and semi-autonomous systems overlap in well-structured environments. An example of a semi-autonomous system is a self-navigating vacuum cleaner that can recognize and maneuver around obstacles.

  • Nearly Full Autonomous Systems can perform many automated tasks, but the automatic functions are still activated or deactivated by an operator. These systems, when activated, can function without the control of an operator but lack some of the adaptability and decision making of a fully autonomous system. An example of a nearly full autonomous system is a self-driving car that can maneuver around obstacles, sense and interpret street signs and traffic lights, and choose the best course to its destination.

  • Fully Autonomous Systems require no human intervention to perform tasks, even in drastically changing environments. A fully autonomous system assesses the environment and adapts to it to complete its mission. An example of a fully autonomous system is a deep-space probe expected to complete its mission without communication from Earth.

An adaptive system can make changes in its performance according to its state, the environment it finds itself in, and a changing mission. Adaptive systems can range from being automatic to fully autonomous. Adaptive systems can be a set of discrete interacting, interdependent, real, or abstract entities that react together to environmental changes or changes in system status. A distributed adaptive system is a system in which an understanding of the individual parts does not necessarily convey an understanding of the whole system's behavior.

The modern battlefield has automated and various levels of autonomous systems working together in a complex, unstructured, operational environment. Future battlefields are expected to have an increased number of autonomous systems working together or independently on shared or independent missions.

This work was done by Jayashree Harikumar and Philip Chan for the Army CCDC Data & Analysis Center. ARL-0219



This Brief includes a Technical Support Package (TSP).
Document cover
Developing Knowledge and Understanding for Autonomous Systems for Analysis and Assessment Events and Campaigns

(reference ARL-0219) is currently available for download from the TSP library.

Don't have an account? Sign up here.