论文部分内容阅读
针对不确定环境下机器人行为控制的维数灾难和感知混淆问题,引入神经元激励机制,提出一种情景记忆驱动的马尔可夫决策过程(EM-MDP)以实现机器人对环境经验自主学习,及多源不确定性条件下的行为控制.首先,构建情景记忆模型,并基于认知神经科学提出事件中状态神经元激活及组织机制.其次,基于自适应共振理论(ART)与稀疏分布记忆(SDM)通过Hebbian规则实现情景记忆的自主学习,采用神经元突触势能建立机器人行为控制策略,机器人能够评估过去的事件序列,预测当前状态并规划期望的行为.最后,实验结果验证,该模型框架与控制策略能够实现机器人在普遍场景中的行为控制目标.
Aimed at the problem of dimensionality disaster and perceived confusion of robot behavior control in uncertain environment, a neuron-based incentive mechanism is introduced and a scenario-memory-driven Markov Decision Process (EM-MDP) is proposed to realize robot’s autonomous learning of environmental experience. Behavioral control under multi-source uncertaintiesFirstly, a scenario memory model was constructed and neuron activation and organization mechanism was proposed based on cognitive neuroscience.Secondly, based on adaptive resonance theory (ART) and sparse distribution memory SDM) to autonomic learning of contextual memory through Hebbian rules, using the synaptic potential of neurons to establish a robot behavior control strategy, the robot can evaluate the past sequence of events, predict the current state and plan the expected behavior.Finally, the experimental results show that the model framework And control strategy to achieve the goal of behavior control of robots in common scenarios.