Human-Machine Communications for Autonomous Systems

Abstract

This paper explores some of the difficult issues associated with the ability of deployed, operational Autonomous Remotely Piloted Airborne Systems (ARPAS) to make decisions and then act upon the decisions made, particularly with regard to the dialogue with the Human Operator or Supervisor. We then look at how the use of operational analysis techniques can extend into Communications Modelling which, when coupled with a Synthetic Environment, allows developers to experiment with Autonomous capabilities against the backdrop of a high-fidelity communications landscape in a safe and secure environment, thus smoothing the path towards eventual integration and deployment

Firstly we explore what an Autonomous System needs in order to act autonomously by considering the question: How does the decision-making software decide what to do and what information does it need in order to do it? We then consider the role of the Human in such systems and how his/her role is becoming more Supervisory in nature, as described by schemes such as PACT (Pilot Authorisation and Control of Tasks) developed by Dstl [1], as the levels of autonomy increase in remotely operated systems.

Secondly we consider the circumstances under which the Human Supervisor may wish to control and/or change the type or level of decisions that the Autonomous Systems are allowed to make and/or act upon without recourse to the human in the loop.

Thirdly we consider some of the physical constraints under which an ARPAS might operate with particular regard to the interaction with the Human Supervisor and how they might be optimized. For example, in dealing with periods of time or geographic locations where communications might be restricted; how to assure that the ARPAS conforms to extant operating rules and how the use of an authorization framework can support the devolution of authority to make and act upon decisions.

Finally this paper examines how Modelling and Simulation techniques, methodologies and technologies can help understand and describe the ARPAS ‘problem space’ and be employed as a key enabler towards gaining Regulatory Authority trust, eventually leading to certification to fly, through a worked example using a Commercial off-The Shelf (COTS) toolset.