home · mobile · calendar · colloquia · 1996-1997 · 

Colloquium - Gaudiano

Application of Biological Neural Networks to Mobile Robot Control and Navigation
Paolo Gaudiano
Department of Cognitive and Neural Systems, Boston University

This talk describes three different neural network models that have been developed recently for the control and navigation of a wheeled mobile robot. Each network is based on models of biological function that have been previously developed in the Department of Cognitive and Neural Systems.

The first model learns to control the movements of a mobile robot without any knowledge of the robot's kinematics. This model learns both the forward and inverse kinematics of the robot, and is able to compensate for various sources of noise in the robot's environment. The other two models complement the first one by enabling the robot to navigate toward stationary or moving targets in an unknown environment that includes stationary or moving obstacles.

One of these models is reactive, meaning that it uses a hard-wired neural network to perform specific evasive maneuvers in response to sensory inputs that represent obstacles in the vicinity of the robot. The other model is adaptive: it learns through operant conditioning to recognize patterns of sensor activations that predict an impending collision; after learning is complete, the model uses the learned information to drive the robot around arbitrary obstacles on its way to a target.

Refreshments will be served immediately before the talk at 3:30pm.
Hosted by Dirk Grunwald.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)