Semi-local 3-D Landmarks
Based on Image Sequences for Navigation in Dynamic Environments
Bernd Krieg-Brückner; Thomas Röfer; Rolf Müller; Axel Lankenau
University of Bremen , Department for Informatics
Duration: 1996 - 2002
Headed by Professor Dr. Bernd Krieg-Brückner, the group "Cognitive Robotics" of the Bremen Institute of Safe Systems at the University of Bremen is developing an intelligent means of transport for elder and handicapped persons - the Bremen Autonomous Wheelchair "Rolland". The power wheelchair Meyra Genius 1.522 is equipped with sensors and a PC. It is both, a scientific platform for experiments within the priority program "Spatial Cognition" of the Deutsche Forschungsgemeinschaft (DFG) as well as a demonstrator for the use of formal methods in the development of safety-critical embedded systems.
In the context of the interdisciplinary priority program, the Bremen project structures the different approaches for modeling navigation behaviors from biology, psychology, artificial intelligence, and cognitive robotics in a hierarchical relationship called the taxonomy of navigation: elementary navigation tac tics are combined to tactical navigation in routes; landmarks in open or enclosed space are compared to routemarks in networks of passages; survey knowledge is introduced on the level of strategic navigation. Here, the goal is the unification of concepts, theories, and terminology to allow coordinated tackling the questions from both sides: psychology and biology give priority to the analysis of the mechanisms of navigation, whereas artificial intelligence and cognitive robotics focus on the synthesis of robust and efficient navigation techniques.
Besides the theoretical work on the basics of navigation, robust techniques are examined that allow the wheelchair - based on simple behaviors as wall-following or turning-into-door - to learn even complex skills. The navigation methods developed are founded on the recognition of routemarks (natural or artificial aids of orientation), either by analyzing the wheelchair’s locomotion, or by means of external sensorics, i.e. cameras or ultrasonic sensors.
As first applications,
a driving assistant and a route assistant were implemented. The former
adapts the speed ordered by the user via the joystick to the current obstacle
situation. Thus the wheelchair gradually slows down when approaching an
obstacle. In addition, the driving assistant supports the user by passing
door-frames and by turning round. The route assistant leaves wheelchair
control to the user, but supports him or her with driving hints as, e.g.,
"next junction to the right". This helps patients who cannot remember ways.
In the next development step, the wheelchair will be enhanced - among other
things - by speech input and output capabilities.
implementation, simulation, application
Areas of Research: artificial intelligence; cognitive robotics
Topics: navigation; motion/ dynamic change
Publications of project cooperations:
D - Route Learning
H - Landmark Usage
O - Reference Systems
Q - Spatial Inference
|mail us your comments or remarks||last updated: March 2003|