Computational modeling of cerebellar model articulation controller / Alireza Jalali

• Main Author: Alireza, Jalali
• Format: Thesis
• Published: 2015
• Subjects: QA75 Electronic computers. Computer science; QA76 Computer software
• Online Access: http://studentsrepo.um.edu.my/6047/4/alireza.pdf; http://studentsrepo.um.edu.my/6047/

The cerebellum has major role in the human motor control to coordinate the motion. The cerebellar model articulation controller is a computational model of the human cerebellum. This research is towards the study of cerebellar model articulation controller (CMAC) and its application to non-linear systems. This model of the CMAC is developed to explore its potential for predictive control of movement. The main limitation of Cerebellar Model Articulation Controller is memory size in application for non-linear systems. The size of memory which used by CMAC depends on input space dimension and input signal quantification step. Therefore, the efficient utilization of the CMAC memory is a crucial issue. Our main aim is to develop an optimal CMAC model which decrease memory size and increase the learning accuracy. To solve the memory size problem of CMAC a model namely Hierarchically Clustered Fuzzy Cerebellar Model Articulation Controller (HCFCMAC) is proposed. The performance of the proposed model is simulate and tested to control robotic arm. The presented simulation results show that proposed model is able to obtain a minimal modelling error and increase the learning accuracy. This study is an examination of the HCFCMAC in biped robot control. It addresses simulations of the cerebellum to control robot swing leg. The proposed method includes a new concept of footstep planning strategy based on the Semi Online Fuzzy Q-learning concept for biped robot control in dynamic environments. The main advantages of proposed approach are that, the computing time is very short and the footstep planning for both predictable and unpredictable obstacle in dynamic environment is operational. It will allow the controller to increase the strength. Another main contribution is on obstacle avoidance strategy for robot in dynamic environment. In this research the mathematical model of kinematics and dynamic of biped robot are described. Our approach is on gait pattern planning and control strategy for biped robot iii stepping over dynamic obstacles. The high–level control used to predict the motion of the robot and the low-level control applied to compute the trajectory of swing leg with operation of HCFCMAC.