PhDSeminar

PhD Seminar inTeleoperation, Perception, & adjustable autonomy: How should we interact with robots?: 

PhD Seminar in Teleoperation, Perception, & adjustable autonomy: How should we interact with robots? Mike Lewis Spring Term 2003

What Should Robots be like?: 

What Should Robots be like? Anthropomorphic? Conversational instructions? General purpose? Autonomous, unsupervised performance? Problem solving? Visual, tactile, proprioceptive, & auditory sensors?

Three Ages of Robotics: 

Three Ages of Robotics Prehistory- 1980’s It’s all in the mind.. AI & planning will make smart robots work (STRIPS & Shakey) 1990’s It’s all in the environment..intelligence & usefulness will just emerge 2000’s It’s all in Human-Robot interaction.. If people and robots could just cooperate we could do great things together..

Shakey’s World(1): 

Shakey’s World(1) The original STRIPS program was designed to control a robot called Shakey. Shakey was capable of moving, grabbing and pushing things, based on plans created by STRIPS.

Knowledge Engineering in Planning Shakey’s world: 

Knowledge Engineering in Planning Shakey’s world Box4 Box3 Box2 Box1 Room1 Room2 Room3 Room4 Corridor Shakey Door1 Door2 Door3 Door4 ls4 ls3 ls2 ls1

Slide6: 

Rodney Brook’s Insect Robots… Fast, Cheap & Out of Control

Slide7: 

One half of HRI is making people want to interact with robots.. Brooks’ COG responds to prosody and attempts to arouse affective responses & engage humans.. Ron Arkin & others have been advising Sony for Aibo (the dog) & their new humanoid robot..

The other half is figuring out how they can cooperate in a way that is:: 

The other half is figuring out how they can cooperate in a way that is: Not completely controlled like an RC car Not completely autonomous like Shakey

Domains I’m Interested in: 

Domains I’m Interested in

Slide11: 

NIST Urban Search And Rescue reference tasks

Common Problems: Navigation: 

Common Problems: Navigation Navigation, context, & situation awareness become difficult when cues to distance, location, etc. are degraded by: narrow field of view absence of cues to depth and distance abrupt changes in field of view uncoupled to controlled motion such as turning a corner

Common Problems: Perception: 

Common Problems: Perception Even significant objects may be difficult to recognize because: narrow field of view absence of cues to depth and distance Unusual/unexpected perspectives Degraded imagery because of nonvisual spectrum sensors, poor quality images, low light levels, etc. Following Examples from Robin Murphy’s Twin Towers Talk

Needed Image Processing and Object Recognition Technologies: 

Needed Image Processing and Object Recognition Technologies -tracks of previous robot run -a watch -3, possibly 4 victims (covered in dust, burned) - head, arm socket?, nose, perhaps fingers WTC 2, 9/18/01

Robots Compared with SearchCams: 

Robots Compared with SearchCams SearchCams ~$10K, Robots ~$12K SearchCams reach up to 5 meters Robots reach up to 30 meters, averaged between 6 and 13 meters Robots can put light on the object, prod it, look at it from different angles Robots can go through more twisting tunnels SearchCams and small robots take <1.5 minutes to set up and insert helped to recognize remains of a body camera with zoom, lights couldn’t do this

State of Available Information: 

State of Available Information what viewpoints have already been explored example: boot? no one rewound tape far enough back to catch earlier viewpoint which disambiguated the object!

Common Problems Autonomy: 

Common Problems Autonomy In both domains we would like to have many robots/person instead of the other way around (like it is today).. To do this we’d like to have robots & humans adjust to each other’s needs.. E.g. Adjustable Autonomy

Seminar Topics: 

Seminar Topics Navigation & situation awareness Egocentric/ exocentric/ & tethered views Augmented Reality (e.g. synthesizing info to help maintain SA) Wayfinding & suggested cues Perception Object recognition/attention direction Control of gaze/ perspective driven locomotion

Topics (continued): 

Topics (continued) Delegation of authority & AA Robot cries for help & human roles in resolving impasses Control of multiple robots..

Planned readings: 

Planned readings Sheridan, Thomas B. (1992). Telerobotics, Automation, and Human Supervisory Control, Cambridge, MA: MIT Press (old but topical) Ali, S. (1999) Multiagent Telerobotics: Matching Systems to Task, Ph.D. Dissertation, Georgia Tech Fong, T.W.(2001) Collaborative Control: A Robot-Centric Model for Vehicle Teleoperation doctoral dissertation, tech. report CMU-RI-TR-01-34, Robotics Institute, Carnegie Mellon University, November, Rogers, E. & Murphy, R. (2001) Final Report forDARPA/NSF Study on Human-Robot Interaction (http://www.aic.nrl.navy.mil/hri/nsfdarpa/) Musliner, D. & Pell, B. (1999) Agents with Adjustable Autonomy: Papers from 1999 AAAI Spring Symposium, AAAI Tech Report SS-99-06 The IJCAI-01 Workshop on Autonomy, Delegation, and Control: Interacting with Autonomous Agents (http://www.csce.uark.edu/~hexmoor/AA01/IJCAI01-cfp.htm)

Project: 

Project Seminar participants will work with game-engine based simulation of USAR reference arenas we are developing to develop and test concepts for teleoperation and/or delegation in control of simulated robot(s)