What You Choose to See Is What You Get: An Experiment with Learnt Sensory Modulation in a Robotic Foraging Task

  • Tiago Rodrigues
  • Miguel Duarte
  • Sancho Oliveira
  • Anders Lyhne Christensen
Conference paper

DOI: 10.1007/978-3-662-45523-4_64

Part of the Lecture Notes in Computer Science book series (LNCS, volume 8602)
Cite this paper as:
Rodrigues T., Duarte M., Oliveira S., Christensen A.L. (2014) What You Choose to See Is What You Get: An Experiment with Learnt Sensory Modulation in a Robotic Foraging Task. In: Esparcia-Alcázar A., Mora A. (eds) Applications of Evolutionary Computation. EvoApplications 2014. Lecture Notes in Computer Science, vol 8602. Springer, Berlin, Heidelberg

Abstract

In evolutionary robotics, the mapping from raw sensory input to neural network input is typically decided by the experimenter or encoded in the genome. Either way, the mapping remains fixed throughout a robot’s lifetime. Inspired by biological sensory organs and the mammalian brain’s capacity for selective attention, we evaluate an alternative approach in which a robot has active, real-time control over the mapping from sensory input to neural network input. We augment the neural controllers with additional output neurons that control key sensory parameters and evolve solutions for a single-robot foraging task. The results show that the capacity to control the mapping from raw input to neural network input is exploited by evolution and leads to novel solutions with higher fitness compared to traditional approaches.

Keywords

Evolutionary robotics dynamic sensors sensor evolution genome-encoding 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Tiago Rodrigues
    • 1
  • Miguel Duarte
    • 1
  • Sancho Oliveira
    • 1
  • Anders Lyhne Christensen
    • 1
  1. 1.Instituto de Telecomunicações & Instituto Universitário de Lisboa (ISCTE-IUL)LisbonPortugal

Personalised recommendations