基于免疫强化学习机制的多机器人动态协作

利用免疫系统中T细胞对抗体的调节作用，本文提出的免疫强化学习方法以强化学习形式，动态调整抗体间相互作用系数，优化网络结构，充分利用了免疫系统的自学习、自适应和免疫记忆特性。将免疫强化学习应用于机器人系统，机器人基于行为集，在实际运行中在线学习未知环境信息，优化行为选择。在基于免疫学习的单机器人系统基础上，考虑多机器人协作性，并应用于多机器人多目标探测中。仿真验证了基于免疫学习机制的多机器人系统对未知动态环境的学习能力和动态协作的有效性。

Multi-robot dynamic cooperation based on an immunized reinforcement learning mechanism

Based on the regulation of antibodies by T-cell in the immune system, the proposed immunized reinforcement learning method employs the reinforcement learning to dynamically adjust the interactions among antibodies, and optimize the network configuration. The learning method fully explores the self-learning, adaptive capability and immune memory of the immune system. Applying the learning method to robotic system, the system learns the unknown environment information and optimizes its behavior employing a behavior-based knowledge and on-line adaptation. Based on the single robot learning, the learning method is applied to the multi-robot observation of multiple moving targets with consideration of multi-robot cooperation. Simulations verify the learning ability in the dynamic unknown environment and the effectiveness of the multi-robot dynamic cooperation based on the learning method.