Vertical motion control of a one legged hopping robot

• Main Author: Azahar, Arman Hadi
• Format: Thesis
• Language: English; English
• Published: 2015
• Subjects: T Technology (General); TJ Mechanical engineering and machinery
• Online Access: http://eprints.utem.edu.my/id/eprint/15860/1/Arman%20Hadi%20bin%20Azahar.pdf; http://eprints.utem.edu.my/id/eprint/15860/2/Vertical%20motion%20control%20of%20a%20one%20legged%20hopping%20robot.pdf; http://eprints.utem.edu.my/id/eprint/15860/; https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=95843
• Institution: Universiti Teknikal Malaysia Melaka

Hopping movement is a desirable locomotion for a mobile robot to adapt on unknown surface and overcome the obstacles avoidance problem. The hopping locomotion is one of locomotion produced by legged robot. The legged type robot has difficult mechanism and complexity in control system. The hopping robot is designed to avoid the obstacles vertically. So, if the hopping robot takes too long time to reach the desired height, it will produced damages to the hopping robot physical. Therefore, the research on develop control strategies of one legged hopping robot is useful so that the developed control strategies can be used and extended to the multi-legged system. Central Pattern Generator (CPG) is a neural network that capable to generate continuous and rhythmic pattern. Since the hopping movement is a continuous and rhythmic jumping movement, it is synthesized that CPG neural network capable to generate hopping movement. Thus, the objectives of this research is to model the one legged hopping robot experimentally, to design a classic controller and integrate with CPG to compensate the steady-state error at each different height, and to optimize the parameters values of Central Pattern Generator (CPG) for the optimum rise time and steady-state error. A hopping peak height detector algorithm is designed to determine hopping peak height as feedback loop. The PI-CPG neural network parameters are optimized for each reference hopping height via simulation. The performance of optimized PI-CPG neural network is evaluated and compared with optimized PI and PID controller. The result shows that the optimized PI-CPG neural network controller produced better response which is 21.36 %, 24.20 %, and 44.13 % average rise time faster than PI-CPG, optimized PI, and optimized PID controller respectively. Moreover, the optimized PI-CPG controller more accurate in term of 4.91 % steady-state error compared to PI-CPG controller; 8.69 %, optimized PI controller; 6.03 %, and optimized PID controller 12.52 % average steady-state error for each reference hopping height. As a conclusion, the hopping height produced by the optimized PI-CPG neural network is more accurate and precise.