Nested Safety Sets for Collision Avoidant Human-Robot Systems. Hawkins, K. & Christensen, H. In
abstract   bibtex   
We investigate the problem of avoiding collision in fast, dynamic human-robot interaction. Optimal control policies are computed for both the human and robot backwards in time from collision. To represent human unpredictability, the robot models the human using a series of increasingly conservative nondeterministic control models. A collection of nested safety sets is found, each of which provides a degree of safety based on how conservative a model the robot must assume of the human to guarantee safety. This allows the robot to anticipate the safety of an interaction based on states labeled as guaranteed safe, guaranteed unsafe, and levels in-between.
@inproceedings {SafePlan16-4,
    track    = {1st Workshop on Planning, Scheduling and Dependability in Safe Human-Robot Interaction (SafePlan 2016)},
    title    = {Nested Safety Sets for Collision Avoidant Human-Robot Systems},
    author   = {Kelsey Hawkins and Henrik Christensen},
    abstract = {We investigate the problem of avoiding collision in fast, dynamic human-robot interaction. Optimal control policies are computed for both the human and robot backwards in time from collision. To represent human unpredictability, the robot models the human using a series of increasingly conservative nondeterministic control models. A collection of nested safety sets is found, each of which provides a degree of safety based on how conservative a model the robot must assume of the human to guarantee safety. This allows the robot to anticipate the safety of an interaction based on states labeled as guaranteed safe, guaranteed unsafe, and levels in-between.},
    keywords = {human-robot safety, collision avoidance, optimal control, differential games, verification}
}
Downloads: 0