Related Experiment Videos
A biologically inspired neural net for trajectory formation and obstacle avoidance
R Glasius1, A Komoda, S C Gielen
1Department of Medical Physics and Biophysics, University of Nijmegen, The Netherlands.
Biological Cybernetics
|June 1, 1996
Summary
This study introduces a novel two-layered neural network for autonomous robots, enabling efficient trajectory formation and obstacle avoidance. The biologically inspired system generates smooth, collision-free paths without external supervision.
Area of Science:
- Computational Neuroscience
- Robotics
- Artificial Intelligence
Background:
- Trajectory formation and obstacle avoidance are critical challenges in autonomous systems.
- Existing methods often require complex control or external supervision.
Purpose of the Study:
- To present a biologically inspired, two-layered neural network for robust trajectory formation and obstacle avoidance.
- To demonstrate a self-organizing system capable of generating smooth, collision-free paths.
Main Methods:
- Developed a two-layered neural network with analog neurons and continuous dynamics.
- Utilized topographically ordered sensory and motor maps.
- Simulated a point robot and a multi-joint manipulator to validate the network's performance.
Main Results:
- The sensory map dynamically updated to reflect optimal paths based on changing target and obstacle positions.
- The motor map generated a moving activity cluster, interpreted as a population vector, guiding movement.
- The system successfully navigated to targets, even under external perturbations, producing smooth trajectories.
Conclusions:
- The proposed neural network effectively achieves trajectory formation and obstacle avoidance through intrinsic network dynamics.
- This biologically inspired model offers a potential solution for direct control of autonomous systems in cluttered environments.
- The system's ability to adapt to dynamic environments and perturbations highlights its robustness.