Mixed-Initiative Control of Robotic Systems

Abstract : A critical topic of research concerning human interaction with robotic warriors concerns the functionality of intelligent systems to advise human operators and share control of robots with those operators. This functionality will engage human and software systems in a complex, highly interdependent exchange of information and control as humans initialize systems that advise them, refine system recommendations, and trade off control of robotic forces with the system during mission execution. In research for DARPA and the U.S. Army, the authors have defined the Relational Knowledge Framework (RKF) that defines fundamental classes of human interactions with intelligent robotics systems planning and control systems. Several cognitive issues are prominent in these interactions. They suggest that system design and training should support specific types of knowledge by operators. These concern the relations (thus, the relational knowledge framework) between the following: (1) the current state of the battle or the system and norms, (2) system parameters and system operations, (3) system inputs and real-world events, and (4) control decisions and the control interface. The framework, cognitive issues, and training and design requirements are defined.