ICANN 1996: Artificial Neural Networks — ICANN 96 pp 311-316 | Cite as
Robot learning in analog neural hardware
Abstract
This paper describes a mobile robot that learned local maneuvers according to the principles of classical and operant conditioning, which were performed in an analog neural hardware implementation. The neurons were equipped with fixed gain and Hebbian inputs, each of which was low-pass filtered for short term memory effects. A wheelchair equipped with sonar and tactile sensors was used as a mobile robot that was able to steer autonomously through narrow doorways after learning an obstacle avoidance task. The system presented here performed operant conditioning in analog hardware controlling a physical mobile robot, which, to our knowledge, was not shown before.
Keywords
Mobile Robot Operant Conditioning Hebbian Learning Steering Angle Sonar SensorPreview
Unable to display preview. Download preview PDF.
References
- Bühlmeier, A. 1996. Analog Neural Networks in Autonomous Systems. Ph.D. thesis, Universität Bremen, Germany. forthcoming.Google Scholar
- Card, H.C., Dolenko, B.K., McNeill, D.K., Schneider, C.R., & Schneider, R.S. 1994. Is VLSI Neural Learning Robust Against Circuit Limitations ? Pages 1889–1893 of: Proc. 1994 IEEE International Conference on Neural Networks.Google Scholar
- Douglas, R., Mahowald, M., & Mead, C. 1995. Neuromorphic Analogue VLSI. Ann. Rev. Neuroscience, 18, 255–281.Google Scholar
- Manteuffel, G. 1992. Neuronal Analog Models with Plastic Synapses for Neurobiological Teaching and Robotics. Pages 155–162 of: Schuster, H.G. (ed), Applications of Neural Networks. VCH Verlagsgesellschaft.Google Scholar
- Mead, C. 1990. Neuromorphic Electronic Systems. Proc. IEEE, 78, 1629–1636.Google Scholar
- Rossmann, M., Hesse, B., Goser, K., Bühlmeier, A., & Manteuffel, G. 1996. Implementation of a Biologically Inspired Neuron-Model in FPGA. Pages 322–329 of: Proceedings of MicroNeuro 96.Google Scholar