用于机器人运动控制的通用小脑认知模块的构建

用具有神经生理学性质的小脑模型实现对机器人系统各种不同感觉协调运动的控制,这是当前机器人控制领域的一个研究热点。但针对一种控制任务就要构建一种对应的小脑模型,这种方式效率很不理想。文中以运动控制为基础,构建了一种包括所有主要细胞类型及其连接的通用小脑认知模块(General Cerebellum CognitiveModule,GCCM)。对给定的背景知识,GCCM能够通过学习生成理想输出,因而,可用于各种不同的机器人感觉运动控制任务。仿真实验证明,通过将脊髓层次上的轨迹误差检测与小脑中的记忆轨迹结合起来,嵌入的GCCM控制系统能对快速手臂延伸运动进行精确的鲁棒控制。

Development of General Cerebellar Cognitive Module Used for Robot Motor Control

The researches on cerebellar model and sensory-motor system have gradually become an important direction in the field of robot technology.But it is not an ideal method to construct a special cerebellar model for every control task.Base on motor control theory,this paper has developed a general cerebellar cognitive module(GCCM) incorporating all major cell types and connections.GCCM learns to produce a desired output for a given context.As such its function is determined by how its inputs and outputs are connected in the system that it is embedded in.As a result,the GCCM can be used for different sensorimotor coordination control tasks in robot system.The simulation results show that the embedded cerebellar model allows stable and robust control of fast reaching movements if it combines detection of the trajectory errors at the spinal cord level with memory traces in the cerebellum.