节律运动和感知信息耦合机理研究及实验验证

Rhythmic movement and sensory feedback cannot be fused effectively, which has become a common problem in CPG-inspired robot walking control. Because of the weak modulation between robots and the information of environments, it is difficult to adjust the motion patterns in real-time according to the walking states of the robot, which leads to the case that it is hard or even impossible for adaptive walking of the robot in the complex environment. In this project, by considering the staged optimization algorithm, we will develop a novel control method via the imitation of layered reception of reflexes and layered execution of the motion commands in biological CPG. The main topics include: (1) build a novel hierarchical CPG model to improve and standardize the existing multi-layered neuron network; (2) by imitating the biological reflex mechanism, propose feedback paths of the multi-sensor information and the strategy of information reception and interconnection; explore the fused strategies when some performance, such as, the real-time quality and computation efficiency, are taken into account in practice; (3) design a novel staged optimization algorithm with variable searching space such that the optimized structure of the control system could be obtained, and develop an intelligent platform which effectively integrates human intelligence and computational intelligence; (4) demonstrate the effectiveness of the proposed methods on environmental walking adaptability of the humanoid robot. In this project, it is aimed to explore novel control theories and approaches for the problem of humanoid walking, which are expected to greatly improve the innovations of biological-inspired control methods and applications in related engineering fields.

节律运动和感知信息无法有效融合是目前中枢模式发生机制（CPG）机器人工程应用的共性问题。传感信息对机器人与环境交互的调节作用较弱，较难根据机器人运动状态实时调整运动方式，无法完成复杂环境中的适应性行走。本研究拟结合进化计算，模拟生物CPG反射信息分层接收和运动命令分层执行的机制来解决这一问题，研究内容包括：(1)建立一种分层CPG模型，完善和规范现有的多层CPG模型；(2)模拟生物反射机制来建立多传感信息的反馈通路，提出能实现多传感信息接收与耦合的策略，探索耦合策略在系统实现面临实时性、计算效率等关键问题时的解决方法；(3)设计变搜索空间的分步式优化算法，使模型能以最优的结构形式应用到工程中。开发一个有效融合人类智能和计算智能的设计平台；(4)进行仿人机器人适应性行走的实验设计与验证。本项目旨在探索仿人机器人行走控制新的理论和方法，能极大地推动生物诱导控制方法在诸多应用领域内的设计创新。

节律运动和感知信息无法有效融合是目前中枢模式发生机制（CPG）机器人工程应用的共性问题。传感信息对机器人与环境交互的调节作用较弱，较难根据机器人运动状态实时调整运动方式，无法完成复杂环境中的适应性行走。本研究拟结合进化计算，模拟生物CPG反射信息分层接收和运动命令分层执行的机制来解决这一问题，研究内容包括：(1)建立一种分层CPG模型，完善和规范现有的多层CPG模型；(2)模拟生物反射机制来建立多传感信息的反馈通路，提出能实现多传感信息接收与耦合的策略，探索耦合策略在系统实现面临实时性、计算效率等关键问题时的解决方法；(3)针对仿人机器人的关节空间行走控制，将CPG与强化学习算法相融合，提出分层强化学习模型；(4)进行仿人机器人适应性行走的实验设计与验证。本项目旨在探索仿人机器人行走控制新的理论和方法，能极大地推动生物诱导控制方法在诸多应用领域内的设计创新。