Neural Networks and Q-Learning for Robotics

International audience IntroductionBehavior-Based ApproachSupervised Learning of a BehaviorMiniature Mobile Robot KheperaIllustration: Reward-Penalty LearningReinforcement LearningGenetic AlgorithmsLearning Classifier SystemsGA & ANNQ-learningEvaluation FunctionAlgorithmReinforcement FunctionUpdate FunctionConvergenceLimitationsGeneralizationNeural Implementations of the Q-learningMultilayer Perceptron Implementation (ideal & Q-CON)Q-KOHONComparisonKnowledge IncorporationReinforcement Function DesignBuilding of a non-explicit ModelLearning in Cooperative RoboticsReferences