Abstract

Introduction

Materials & Methods

Results

Discussion & Conclusion

References

Discussion
Board

The problem of action selection is to decide on line ’what to do next'. It has long been studied in animals by ethologists (McFarland, 1977), in humans by psychologists (Rachlin, 1989) and, more recently, in robots by computer scientists (Tyrrell,1993). Our work is centered on the strategies of action selection in laboratory mice (C3H strain), with the goal of implementing them in biologically inspired autonomous robots.

Two action selection strategies have been already identified by classical linear analyses: The first (System I) concerns the choice of ultradian alternations of rest and activity bouts, and the second (System II) concerns the choice of the succession of acts in each activity bout. System I seems to allow mice to get a global high net energy gain (energy input minus energy output). System II has a less clear functional role, as it leads to great interindividual variabilities, some mice increasing, but some other decreasing, their net energy gain while performing the succession of acts in the activity bouts.

System I exhibits periodic dynamics (Guillot & Meyer, 1995). The purpose of this paper is to determine, by nonlinear analysis, what kind of dynamics - stochastic or nonlinear deterministic - is generated by System II, considering that it is not periodic (Guillot & Meyer, 1997). Our hypothesis is that it may be chaotic.

Back to the top.