"TROLL": a regenerating robot

From ISLAB/CAISR
Title "TROLL": a regenerating robot
Summary A robot which can detect faults on itself and try to mark or fix them
Keywords regenerative robot, self-detection/body schema, anomaly/fault detection, self-healing/augmentation
TimeFrame 2015-01-16 - 2015-06-31
References (self-detection) K. Gold, B. Scassellati, Using probabilistic reasoning over time to self-recognize, Robotics and Autonomous Systems (2008), doi:10.1016/j.robot.2008.07.006

(body schema) Mai Hikita, Sawa Fuke, Masaki Ogino, Takashi Minato and Minoru Asada. Visual attention by saliency leads cross-modal body representation. IROS - 2008. (anomaly detection) Takahiro Suzuki, Fumihiro Bessho, Tatsuya Harada and Yasuo Kuniyoshi. Visual Anomaly Detection under Temporal and Spatial Non-uniformity for News Finding Robot. IROS 2011. (self-augmentation) Luzius Brodbeck and Fumiya Iida. Enhanced Robotic Body Extension with Modular Units, IROS 2012.

Prerequisites some capability to work with software, and interest in robots and recognition
Author Yinrong Ma
Supervisor Martin Cooney, Anita Sant'Anna
Level Master
Status Finished


Goal: a companion robot which can fix some of its own problems itself

Motivation: people, especially the elderly, cannot be expected to maintain very complex robots; robots should help out

Challenge: it's tough even for humans to repair robots. For a soft companion robot intended to interact closely with people, two very common problems are:

 Degradation of its exterior (marks and scratches)
 Motor problems (due to e.g., frayed/loose cables, gears wearing down)

Approach:

 To detect the problems above, the student will set up a robot to look at itself periodically 
   in a mirror while moving, comparing a mental model of itself with what it actually perceives. 
 Time allowing, the student should also try to develop some capability for the robot to seek to 
   repair identified problems: marking holes in its exterior and damaged motors, and possibly 
   robot motions to try to fill holes with foam

Requirement: some capability to work with software, and interest in robots and recognition

Expected results: a thesis/report, code, video