General document | Modelling Human Attributes in Autonomous Systems Operations
Unmanned Systems (UMS) have the potential to provide significant benefits when used in military operations that are dangerous, dirty, and/or dull; the principal benefit cited in a variety of literature is of reducing the risk of human casualties. Other claimed benefits of UMS suggest they can enhance situational awareness, reduce human workload, and improve mission performance at a reduced cost. These are worthwhile benefits; however, war is a particularly human endeavour in which current technology cannot replicate the understanding, intuition and decision making of human combatants. Human-UMS teaming is seen as a desirable state in which humans supervise, task and interact with robotic systems. This is the semi-autonomous mode of operation of UMS, also known as human supervisory controlled (HSC) autonomy; for example, the USAF fleet of long endurance Unmanned Aerial Vehicles (UAVs). In HSC UMS military employment, there is evidence from simulator-based experiments that the operators’ attributes have an impact on the UMS performance. However, it is recognised that while the autonomous systems are modelled with high fidelity, the human element is poorly represented for the human-operated systems in extant closed-loop combat simulation models.
Therefore these simulations, when applied to model HSC UMS employment, can lack proper characterisation of human attributes. In this report we review the literature to investigate examples where human-attributes have been modelled in other simulation paradigms, system dynamics (SD) and discrete event simulation (DES) in particular.
More specifically, based on our review, we have identified five human attributes (trust, impact of human interventions, cognitive workload, attention allocation and situation awareness, and human learning) that have shown to have an impact on UMS system performance. We discuss how each of the identified human attributes is implemented, and contributes to the modelling and measurement of system performance in the SD and DES models.
The differences of SD and DES approaches are compared and analysed. Lessons learnt from SD or DES models are described. These lessons could to inform combat simulation practitioners to enrich the existing high-fidelity combat simulation models.