AUTHOR=Cyr André , Boukadoum Mounir , Thériault Frédéric 
  
TITLE=Operant conditioning: a minimal components requirement in artificial spiking neurons designed for bio-inspired robot's controller
  
JOURNAL=Frontiers in Neurorobotics
  
VOLUME=Volume 8 - 2014
  
YEAR=2014
  
URL=https://www.frontiersin.org/journals/neurorobotics/articles/10.3389/fnbot.2014.00021
  
DOI=10.3389/fnbot.2014.00021
  
ISSN=1662-5218
  
ABSTRACT=We demonstrate the operant conditioning (OC) learning process within a basic bio-inspired robot controller paradigm, using an artificial spiking neural network (ASNN) with minimal component count as artificial brain. In biological agents, OC results in behavioral changes that are learned from the consequences of previous actions, using progressive prediction adjustment triggered by reinforcers. In a robotics context, virtual and physical robots may benefit from a similar learning skill when facing unknown environments with no supervision. In this work, we demonstrate that a simple ASNN can efficiently realise many OC scenarios. The elementary learning kernel that we describe relies on a few critical neurons, synaptic links and the integration of habituation and spike-timing dependent plasticity (STDP) as learning rules. Using four tasks of incremental complexity, our experimental results show that such minimal neural component set may be sufficient to implement many OC procedures. Hence, with the described bio-inspired module, OC can be implemented in a wide range of robot controllers, including those with limited computational resources.