A bionic cerebellar motion control model and its application in arm control

在小脑皮层神经元水平上研究小脑如何实现对肢体运动的控制并运用到智能机器人系统中，是目前人工智能和康复医学等研究领域的一个热点。目前通常使用的小脑模型仅以控制效果为目的，虽借鉴了小脑的功能模式，却忽略了小脑的结构特性。实际上，小脑模型除了用于实现控制目的以外，还应该具有控制过程的可解释性并能分析小脑发生病变时带来的后果等，所以需要建立一种更能表达小脑特性的仿生小脑模型。本文在神经元水平上探讨了小脑如何处理外部输入信息进而产生控制指令的过程，通过对具有匀质结构的小脑进行功能化分块，构建了一种包含小脑皮层主要细胞类型和细胞间连接方式的新型仿生小脑运动控制模型。通过仿真实验和力反馈器控制实验表明，本文构建的仿生小脑运动控制模型与目前广泛应用的小脑关联控制器模型相比，具有更好的控制效果，从而验证了本文仿生小脑运动控制模型的有效性，为进一步实现真正意义上的类脑人工智能控制奠定了基础。.; How to realize the control of limb movement and apply it to intelligent robot systems at the level of cerebellar cortical neurons is a hot topic in the fields of artificial intelligence and rehabilitation medicine. At present, the cerebellar model usually used is only for the purpose of controlling the effect, borrowing from the functional mode of the cerebellum, but it ignores the structural characteristics of the cerebellum. In fact, in addition to being used for controlling purposes, the cerebellar model should also have the interpretability of the control process and be able to analyze the consequences of cerebellar lesions. Therefore, it is necessary to establish a bionic cerebellar model which could better express the characteristics of the cerebellum. In this paper, the process that the cerebellum processes external input information and then generates control instructions at the neuron level was explored. By functionally segmenting the cerebellum into homogeneous structures, a novel bionic cerebellar motion control model incorporating all major cell types and connections was established. Simulation experiments and force feedback device control experiments show that the bionic cerebellar motion control model can achieve better control effect than the currently widely used cerebellar model articulation controller, which verifies the effectiveness of the bionic cerebellar motion control model. It has laid the foundation for real brain-like artificial intelligence control.