Cue verification for an automatically-guided wheelchair system

An automatically-guided wheelchair system has been developed at the University of Notre Dame. Like some other automatically-guided vehicles (AGVs), this system uses filtering to combine information from measurements and the state equations to produce accurate ongoing pose estimates. This approach relies on the detection of nonunique passive visual cues placed at known positions throughout the environment. In order to maintain accurate pose estimates, a measurement involving any given detected cue must correspond with the correct cue identity. This paper deals with the approach taken to address the problem of establishing correct identity