Author name cluster

Marcel Schoppers

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

2 author rows

ICRA Conference 2001 Conference Paper

Stereo Ego-motion Improvements for Robust Rover Navigation

Clark F. Olson
Larry H. Matthies
Marcel Schoppers
Mark W. Maimone

Robust navigation for mobile robots over long distances requires an accurate method for tracking the robot position in the environment. Techniques for position estimation by determining the camera ego-motion from monocular or stereo sequences have been previously described. However, long-distance navigation requires a very high level of robustness and a very low rate of error growth. In this paper, we describe a methodology for long-distance rover navigation that meets these goals using robust estimation. We show that a system based on only camera ego-motion estimates will accumulate errors with super-linear growth in the distance travelled, owing to increasing orientation errors. When an absolute orientation sensor is incorporated, the error growth can be reduced to a linear function of the distance travelled. We tested these techniques using both extensive simulation and hundreds of real rover images and achieved a low, linear rate of error growth.

Details

AAAI Conference 1996 Conference Paper

Generalizing Indexical-Functional Reference

Marcel Schoppers

The goals of situated agents generally do not specify particular objects: they require only that some suitable object should be chosen and manipulated (e. g. any red block). Situated agents engaged in deictic reference grounding, however, may well track a chosen referent object with such fixity of purpose that an unchosen object may be regarded as an obstacle even though it satisfies the agent’ s goals. In earlier work this problem was bridged by hand-coding. This paper lifts the problem to the symbol level, endowing agents with perceptual referent selection actions and performing those actions as required to allow or disallow opportunistic re-selection of referents. Our work preserves the ability of situated agents to find and track specific objects, adds an ability to automatically exploit the opportunities allowed by nonspecific references, and provides a starting point for studying how much opportunistic perception is appropriate.

PDF Details

AIJ Journal 1995 Journal Article

The use of dynamics in an intelligent controller for a space faring rescue robot

Marcel Schoppers

The NASA Extra Vehicular Activity Retriever (EVAR) robot is being designed to retrieve astronauts or objects that become detached from the orbiting Space Station. This task requires that the robot's intelligent controller must rely heavily on orbital dynamics predictions, without becoming blind to the wide variety of anomalies that may occur. This article describes the controller's Universal Plan (U. P.) and some technical lessons learned from it. The U. P. reacts not to actual current states but to estimated states, which are obtained using goal-directed active perception. A modal logic formalization of discrete-event dynamics allows us to finely analyze and specify the interactions of knowledge, belief, sensing, acting, and time within the U. P. The U. P. now acts like a hands-off manager: it makes regular observations, grants some leeway for unobservable or ill-modelled processes, has faith in subsystem dynamics, and takes action only to manipulate subsystems into delivering desired progress. Most of the time, the appropriate action is to do nothing. Finally we examine properties of the application that allowed the U. P. to deliver robust goal achievement despite misleading state estimates, weak models of relevant processes, and unpredictable disturbances.

Details DOI

AAAI Conference 1994 Conference Paper

Estimating Reaction Plan Size

Marcel Schoppers

The Shannon/Ginsberg circuit size estimate, by assuming independence of Boolean inputs, is not usable as a plan size estimate. By re-estimating circuit size as a function of the number of combinations w of Boolean inputs, I show that a reaction plan over w world states should grow as O(w/log w), on average. Finally I obtain the general domain-independent result that for a domain containing w world states, the expected size of a reaction plan with variables is O.

PDF Details