2010.69: A Comparison of Multivariable & Decentralized Control Strategies for Robust Humanoid Walking
2010.69: H. Dallali, G.A. Medrano-Cerda and M. Brown (2010) A Comparison of Multivariable & Decentralized Control Strategies for Robust Humanoid Walking. In: UKACC International Conference on CONTROL 2010, 7-10 Sept 2010, Coventry, UK.
Full text available as:
|PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader|
Official URL: http://vs-ec-ctacconf.coventry.ac.uk/drupal-6.14/
Bipedal walking is one of the most interesting control problems in humanoids research. Walking is modelled as a hybrid system in the sense that it involves various phases such as single support phase, impacts with the ground (i.e. a state reset) and the double support phase. The control system has to provide good dynamic performance in these different modes to achieve fast walking speeds while guaranteeing its safe and robust operation. Most humanoids use local joint PID loops (decentralized) control systems while the robot is a multivariable system and walking involves significant interactions between the robot links. Hence in this paper a centralized LQR multivariable controller is designed for the robot and analyzed for its stability, robustness to noise and disturbances and dynamic performance. Then, an LQR based iterative algorithm is used to tune the local PID servos. A comparison between the two schemes is done, where it is shown that the multivariable LQR has better robustness and energy efficiency. Finally, both controllers are simulated using the linearized model of a 10 degree of freedom robot called “C-Cub”.
|Item Type:||Conference or Workshop Item (Paper)|
|Uncontrolled Keywords:||CICADA, Linear multivariable control, Decentralized control, Humanoid locomotion, C-Cub|
|Subjects:||MSC 2000 > 15 Linear and multilinear algebra; matrix theory|
MSC 2000 > 34 Ordinary differential equations
MSC 2000 > 93 Systems theory; control
|Deposited By:||Mr Houman Dallali|
|Deposited On:||18 August 2010|