Generalized Shared Control versus Classical Shared Control: Illustrative Examples

Figure 1: The autonomy (black solid lines) can take 2 possible routes through the obstacle field (thus displaying multimodality), both of which have nearly the same safety and efficiency. The operator model (red dashed lines) indicates two potential paths forward (such multimodality might arise as a result of unreliable communications, a multitasking operator, or a startled operator). Linear blending (purple dashed arrow) blends the “most likely operator future” on the left with the “most likely autonomy future” on the right, resulting in an operator-disagreeable solution and a possible collision. IRT (brown solid arrow) can blend either of the two operator futures with either of the two autonomy futures. In this case, the data drives IRT to blend the “most likely operator future” with “autonomy future 2”, resulting in a solution that is safe and in agreement with the operator. Shared control fuses operator inputs and autonomy inputs into a single command. However, if environmental or operator predictions are multimodal, state of the art approaches are suboptimal with respect to safety, efficiency, and operator-autonomy agreement: even under mildly challenging conditions, existing approaches can fuse two safe inputs into an unsafe shared control. Multimodal conditions are common to many real world applications, such as search and rescue robots navigating disaster zones, teleoperated robots facing communication degradation, and assistive driving technologies. In [11], we introduced a novel approach called interacting random trajectories (IRT) that simultaneously optimizes autonomy objectives (e.g., safety and efficiency) and operator-autonomy agreement under multimodal conditions; this optimality prevents such unsafe shared control. In this paper, we describe those results in more user friendly language by using illustrations and text.