SVR control of the sagittal balance of a biped robot combining torso and ankle joints

This paper describes the control of an autonomous biped robot that combines the use of the torso and the ankle joints movements for its sagittal balance. The innovative characteristic of this controller is the combined use of the ankle and torso joints movements to correct the Zero Moment Point (ZMP). It is used an artificial intelligence technique, the Support Vector Regression (SVR), to control the balance of the robot. The relation between torso and ankle compensations was obtained using the traditional inverted pendulum method. The use of the ankle joint for balance control implies the real time change of the designed initial gait but presents a lower angle control variation compared to the use of the torso joint which is simpler and more precise but slower and with a lower ZMP compensation range. Combining the two control methods it is possible to achieve better results. To obtain a good stable step it is very important to have a good initial legs trajectory design. With that purpose in mind human-based trajectories were used, leading to smaller control corrections of ankle and torso joints. It were tested different combinations of torso and ankle joints corrections for the balance control on flat horizontal and inclined surfaces and the results presented. In order to evaluate and compare the performance of the balance control methods of a biped robot two performance indexes are proposed.