Utilising a cerebellar model for mobile robot control in a delayed sensory environment

Fast and accurate control of a system exhibiting significant feedback delay is traditionally a difficult problem to solve. In biological systems, it is thought that a part of the brain called the cerebellum overcomes such difficulties. This paper outlines the use of a cerebellar model in the control of a simulated mobile robot. The model is based around Albus’s CMAC neural network, and uses the response of a non-delayed teaching module as a basis for learning. The model was able to produce results comparable to the teacher despite being subjected to severe sensory latency. After limited initial training the system can rapidly adapt to new situations and proved to have good generalisation between similar movements.