Timing without Time — An Experiment in Evolutionary Robotics

  • H. H. Lund
Conference paper


Hybrids of genetic algorithms and artificial neural networks can be used successfully in many robotics applications. The approach to this is known as evolutionary robotics. Evolutionary robotics is advantageous because it gives a semi-automatic procedure to the development of a task-fulfilling control system for real robots. It is disadvantageous to some extent because of its great time consumption. Here, I will show how the time consumption can be reduced dramatically by using a simulator before transferring the evolved neural network control systems to the real robot. Secondly, the time consumption is reduced by realizing what are the sufficient neural network controllers for specific tasks. It is shown in an evolutionary robotics experiment with the Khepera robot, that a simple 2 layer feedforward neural network is sufficient to solve a robotics task that seemingly would demand encoding of time, for example in the form of recurrent connections or time input. The evolved neural network controllers are sufficient for exploration and homing behaviour with a very exact timing, even though the robot (controller) has no knowledge about time itself.


Real Robot Neural Network Controller Simulated Robot Recurrent Connection Evolutionary Robotic 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    W.-P. Lee, J. Hallam, and H. H. Lund. A Hybrid GP/GA Approach for Co-evolving Controllers and Robot Bodies to Achieve Fitness-Specified Tasks. In Proceedings of IEEE Third International Conference on Evolutionary Computation, NJ, 1996. IEEE Press.Google Scholar
  2. [2]
    H. H. Lund and J. Hallam. Sufficient Neurocontrollers can be Surprisingly Simple. Research Paper 824, Department of Artificial Intelligence, University of Edinburgh, 1996.Google Scholar
  3. [3]
    H. H. Lund, J. Hallam, and W.-P. Lee. Evolving Robot Morphology. In Proceedings of IEEE Fourth International Conference on Evolutionary Computation, NJ, 1997. IEEE Press. Invited paper.Google Scholar
  4. [4]
    H. H. Lund and O. Miglino. Prom Simulated to Real Robots. In Proceedings of IEEE Third International Conference on Evolutionary Computation, NJ, 1996. IEEE Press.Google Scholar
  5. [5]
    O. Miglino, H. H. Lund, and S. Nolfi. Evolving Mobile Robots in Simulated and Real Environments. Artificial Life, 2(4):417–434, 1996.CrossRefGoogle Scholar
  6. [6]
    O. Miglino, K. Nafasi, and C. Taylor. Selection for Wandering Behavior in a Small Robot. Artificial Life, 2(1), 1995.Google Scholar
  7. [7]
    F. Mondada, E. Pranzi, and P. Ienne. Mobile robot miniaturisation: A tool for investigation in control algorithms. In Experimental Robotics III. Lecture Notes in Control and Information Sciences 200, pages 501–513, Heidelberg, 1994. Springer-Verlag.Google Scholar

Copyright information

© Springer-Verlag Wien 1998

Authors and Affiliations

  • H. H. Lund
    • 1
  1. 1.Department of Artificial IntelligenceUniversity of EdinburghEdinburghScotland, UK

Personalised recommendations