Shared control is an increasingly popular approach to facilitate control and communication between humans and intelligent machines. However, there is little consensus in guidelines for design and evaluation of shared control, or even in a definition of what constitutes shared control. This lack of consensus complicates cross fertilization of shared control research between different application domains. This paper provides a definition for shared control in context with previous definitions, and a set of general axioms for design and evaluation of shared control solutions. The utility of the definition and axioms are demonstrated by applying them to four application domains: automotive, robot-assisted surgery, brain–machine interfaces, and learning. Literature is discussed for each of these four domains in light of the proposed definition and axioms. Finally, to facilitate design choices for other applications, we propose a hierarchical framework for shared control that links the shared control literature with traded control, co-operative control, and other human–automation interaction methods. Future work should reveal the generalizability and utility of the proposed shared control framework in designing useful, safe, and comfortable interaction between humans and intelligent machines.

@en

When using an automated system, user trust in the automation is an important factor influencing performance. Prior studies have analyzed trust during supervisory control of automation, and how trust influences reliance: the behavioral correlate of trust. Here, we investigated how reliance on haptic assistance affects performance during shared control with an automated system. Subjects made reaches towards a hidden target using a visual cue and haptic cue (assistance from the automation). We sought to influence reliance by changing the variability of trial-by-trial random errors in the haptic assistance. Reliance was quantified in terms of the subject's position at the end of the reach relative to the two cues. Our results show that subjects aimed more towards the visual cue when the variability of the haptic cue errors increased, resembling cue weighting behavior. Similar behavior was observed both when subjects had explicit knowledge about the haptic cue error variability, as well as when they had only implicit knowledge (from experience). However, the group with explicit knowledge was able to more quickly adapt their reliance on the haptic assistance. The method we introduce here provides a quantitative way to study user reliance on the information provided by automated systems with shared control.@en

To effectively interpret and interact with the world, humans weight redundant estimates from different sensory cues to form one coherent, integrated estimate. Recent advancements in physical assistance systems, where guiding forces are computed by an intelligent agent, enable the presentation of augmented cues. It is unknown, however, if cue weighting can be extended to augmented cues. Previous research has shown that cue weighting is determined by the reliability (inversely related to uncertainty) of cues within a trial, yet augmented cues may also be affected by errors that vary over trials. In this study, we investigate whether people can learn to appropriately weight a haptic cue from an intelligent assistance system based on its error history. Subjects held a haptic device and reached to a hidden target using a visual (Gaussian distributed dots) and haptic (force channel) cue. The error of the augmented haptic cue varied from trial to trial based on a Gaussian distribution. Subjects learned to estimate the target location by weighting the visual and augmented haptic cues based on their perceptual uncertainty and experienced errors. With both cues available, subjects were able to find the target with an improved or equal performance compared to what was possible with one cue alone. Our results show that the brain can learn to reweight augmented cues from intelligent agents, akin to previous observations of the reweighting of naturally occurring cues. In addition, these results suggest that the weighting of a cue is not only affected by its within-trial reliability but also the history of errors.@en