An Affective-Computing Control Paradigm for Autonomous Systems

Unmanned Ground Vehicles (UGVs) and Unmanned Surface Vehicles (USVs) are assuming an increasingly important role in military operations. They provide effective platforms for performing missions while reducing risk to military personnel. Ultimately, the level of success achieved by UGVs and USVs will be dependent upon the ability to develop autonomous behaviors that are on par with human behaviors. Behavior of current autonomous systems is based on hundreds (or thousands) of programmed parameters. Essentially, these parameters control how a mission is achieved, but are too complex to be easily manipulated in a field environment. The challenge is that the parameters that control the behaviors are buried into a mass of software particularities. Therefore, we propose to develop a series of human understandable abstractions that allow a human operator to easily control how an unmanned system completes a mission and to what extent it uses its capabilities. Example abstractions include aggressiveness, stealth, inquisitiveness, urgency, safety, success and unpredictability (each of these will be further explained in the work plan). It is anticipated that these would be implemented as a set of sliders that allows an operator to control the level of each trait.