Freezing Robot Problem

A classic failure mode of autonomous robots operating among people, first characterised by Trautman and Krause (2010), in which a robot stalls indefinitely because every candidate path is blocked by predicted human motion. The problem arises because overly conservative motion planners expand pedestrian trajectories into uncertainty cones that cover the entire free space, leaving no feasible route. In accessible navigation robots that guide blind users, the freezing-robot problem is a concrete accessibility issue: a robot that stops in a crowded museum, transit hub, or pavement cannot simply ask the user to look around and pick a path. Solutions include modelling human-robot joint collision avoidance, legible motion, and shared-control interactions that let the user (or the robot) request help from bystanders to unblock the path.