Research On Self-Learning Control Methods For Biped Robots Based On Energy-Efficiency Optimization

Since biped robots try to copy locomotions from human being, it suits to help or replace people in their work. Biped walking robots is high order, strong coupling and nonlinear dynamic system with nonholonomic constraints and multiple degrees of freedom. Research on control methods of biped walking robots has not only important academic value but also considerable significance for application. It is related to a diverse of research disciplines to realize stable flexible and intelligent dynamic-walking with high energy efficiency. This dissertation studied some hotspots and difficulties of them, such as energy efficiency and learning control of biped walking.The main line of this dissertation is as follows. Research background and significance are stated in chapter1. Then, chapter2introduces the background of biped robots. Chapter3and chapter4propose detailed strategies for energy-efficiency optimization of biped walking. Self-learning control methods for biped walking robots are presented in chapter5, chapter6and chapter7.The specific researches include:1. A novel systematic methodology of gait control based on energy-efficiency optimization is proposed for solving the fatal problem of high energy consumption in practical application of biped robots. A strategy of energy consumption estimation and an algorithm of energy-efficiency optimization are proposed based on three important indexes of energy consumption for biped locomotion:average mechanical power, mean power derivation and mean torque consumption. Controlling the gait of the robot with the trunk trajectory which corresponds to the minimal energy consumption in the zero moment point (ZMP) stability domain, we obtain the energy-efficiency gait which is guaranteeing the ZMP criterion. The proposed method is applied to a robotic system. Simulation results show the validity of the method compared with other inteligent methods.2. An energy efficient support vector machine (EE-SVM) learning control system considering energy cost of each training sample of biped dynamic is proposed to realize energy efficient biped walking. Energy costs of the biped walking samples are calculated. Then the samples are weighed with the inverses of the energy costs. An EE-SVM objective function with energy related slack variables is proposed, which follows the principle that the sample with the lowest energy consumption is treated as the most important one in the training. That means the samples with lower energy consumption contribute more to the EE-SVM regression function learning, which highly increases the energy efficiency of the biped walking. Simulation results demonstrate the effectiveness of the proposed method.3. An unscented Kalman filter (UKF)-based predictable support vector regression (SVR) learning controller is proposed to improve the flexibility of biped walking robots. After estimating the biped states of the next moment using an UKF, a SVR learning controller with the predicted biped states is implemented to ensure the ZMP stability. Using the predicted biped states, the SVR learning controller can predictably adjust the posture of the trunk timely and properly to adapt to the dynamic posture of the whole body. Simulation and experimental results demonstrate the superiority of the proposed methods.4. To learn biped walking dynamic accurately, and then, compensate time-varying external disturbances timely, a time-sequence-based fuzzy SVM (TSF-SVM) learning control system considering time properties of biped walking samples is proposed. For the first time, time-sequence-based triangular and Gaussian fuzzy membership functions are proposed for the single support phase (SSP) and the double support phase (DSP) respectively according to time properties of different biped phases, which provides an effective way to formulate time properties of biped walking samples in the context of time-varying external disturbances. The performance of the proposed TSF-SVM are compared with other typical intelligent methods, simulation results demonstrate the proposed method is more sensitive to occasional external disturbances, which increases the stability margin and prevents the robot from falling down.5. An interval type-2fuzzy weighted support vector machine (IT2FW-SVM) is proposed to address the problem of high energy consumption for biped walking robots. Different from the traditional machine learning method of 'copy learning', the proposed IT2FW-SVM obtains lower energy cost and larger ZMP stability margin using a novel strategy of'selective learning', which is similar as human selections based on experience. To handle the uncertainty of the experience, the learning weights in the IT2FW-SVM are deduced using an interval type-2fuzzy logic system (IT2FLS), which is an extension of the previous weighted SVM. Simulation studies show that the existing biped walking which generates the original walking samples is improved remarkably in terms of both energy efficiency and biped dynamic balance using the proposed IT2FW-SVM.