Robustness in the Long Run: Auto-teaching vs Anticipation in Evolutionary Robotics

  • Nicolas Godzik
  • Marc Schoenauer
  • Michèle Sebag
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3242)

Abstract

In Evolutionary Robotics, auto-teaching networks, neural networks that modify their own weights during the life-time of the robot, have been shown to be powerful architectures to develop adaptive controllers. Unfortunately, when run for a longer period of time than that used during evolution, the long-term behavior of such networks can become unpredictable. This paper gives an example of such dangerous behavior, and proposes an alternative solution based on anticipation: as in auto-teaching networks, a secondary network is evolved, but its outputs try to predict the next state of the robot sensors. The weights of the action network are adjusted using some back-propagation procedure based on the errors made by the anticipatory network. First results – in simulated environments – show a tremendous increase in robustness of the long-term behavior of the controller.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Meyer, C., Akoulchina, I., Ganascia, J.-G.: Learning Strategies in Games by Anticipation. In: Proc. IJCAI 1997, Morgan Kaufmann, San Francisco (1997)Google Scholar
  2. 2.
    Endo, Y., Arkin, R.C.: Anticipatory Robot Navigation by Simultaneously Localizing and Building a Cognitive Map. In: ICOS 2003 – Intl. Conf. on Intelligent Robots and Systems. IEEE/RSJ (2003)Google Scholar
  3. 3.
    Harvey, I.: Is there another new factor in evolution? Evolutionary Computation 4(3), 311–327 (1997); Special Issue: 100 Years of the Baldwin EffectGoogle Scholar
  4. 4.
    Jordan, M., Rumelhart, D.: Forward models: Supervised learning with a distal teacher. Cognitive Science 16 (1992)Google Scholar
  5. 5.
    Nolfi, S.: How learning and evolution interact: The case of a learning task which differs from the evolutionary task. Adaptive Behavior 7(2), 231–236 (2000)CrossRefGoogle Scholar
  6. 6.
    Nolfi, S., Elman, J.L., Parisi, D.: Learning and evolution in neural networks. Adaptive Behavior 3(1), 5–28 (1994)CrossRefGoogle Scholar
  7. 7.
    Nolfi, S., Floreano, D.: How co-evolution can enhance the adaptive power of artificial evolution: implications for evolutionary robotics. In: Husbands, P., Meyer, J.A. (eds.) Proceedings of EvoRobot 1998, pp. 22–38. Springer, Heidelberg (1998)Google Scholar
  8. 8.
    Nolfi, S., Parisi, D.: Desired answers do not correspond to good teaching input in ecological neural networks. Neural Processing Letters 2(1), 1–4 (1994)CrossRefGoogle Scholar
  9. 9.
    Nolfi, S., Parisi, D.: Learning to adapt to changing environments in evolving neural networks. Adaptive Behavior 5(1), 75–98 (1997)CrossRefGoogle Scholar
  10. 10.
    Kevin O’Regan, J., Noë, A.: A sensorimotor account of vision and visual consciousness. Behavioral and Brain Sciences 24(5) (2001)Google Scholar
  11. 11.
    Stolzmann, W.: An introduction to anticipatory classifier systems. In: Lanzi, P.L., Stolzmann, W., Wilson, S.W. (eds.) IWLCS 1999. LNCS (LNAI), vol. 1813, pp. 175–194. Springer, Heidelberg (2000)CrossRefGoogle Scholar
  12. 12.
    Tishby, N., Pereira, F.C., Bialek, W.: The information bottleneck method. In: Proc. 37th Allerton Conf. on Communication, Control and Computing, pp. 368–377 (1999)Google Scholar
  13. 13.
    Veloso, M., Stone, P., Bowling, M.: Anticipation as a key for collaboration in a team of agents: A case study in robotic soccer. In: SPIE Sensor Fusion and Decentralized Control in Robotic Systems II, vol. 3839 (1999)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Nicolas Godzik
    • 1
  • Marc Schoenauer
    • 1
  • Michèle Sebag
    • 1
  1. 1.TAO teamINRIA Futurs and LRI, UMR CNRS 8623Orsay CedexFrance

Personalised recommendations