Study On Humanoid Online Motion Planning

Study on humanoid robot is one of the most active fields in robotics recently due to its unique motion pattern, and humanoid motion control is the most fundamental and challenging problem undoubtedly.Planning based method is a general way for humanoid motion control. After several years' research, stable and fast humanoid walking on plane, slope and stairs has been realized by using off-line motion planning based method. But it seems still difficult to adapt the robot to uncertain surroundings, which makes it hard to expand application of humanoid robots. Online motion planning based method can combine humanoid dynamics with environmental information in real-time so that it has greater potential to increase the humanoid motion adaptabilty on irregular terrain. So an online motion planning method is proposed for humanoid robot "Blackmann", and some related techniques are studied, including stability estimation, inverse kinematics calculation, disturbance rejection and swing foot landing control. The focus of this dissertation is to find a way to obtain a feasible humanoid walking motion sequence online, but the optimality on some certain index is not mainly concerned.Firstly, the mechanical configuration, control system structure and sensors of the humanoid robot "Blackmann" are introduced, and forward kinematic model using D-H coordinates and dynamical model during single support period using Lagrange method are built. Using VR tools in Matlab, "Virtual Blackmann" is constructed as a humanoid simulation platform, and the flow of inverse dynamics calculation is addressed for validation of the feasibility of planned results.Secondly, after synthesis of traditional humanoid stability judgements, the Extended CoP based judgement and strategy for motion control are presented in virtue of the concept of virtual plane, in order to estimate the humanoid postural stability on uneven terrain during double support period. And strategies of application based on Extended CoP in motion planning are discussed.Thirdly, after summarizing current motion planning methods, a basic variables set based humanoid online motion planning method is proposed. Here, the basic variables set, including Extended CoP, CoG, CoC, ankle point position, posture of coxa and feet, is defined to describe humanoid motion sequences. Thus, the humanoid motion planning is transformed into the planning of the basic variables using mass-concentrated model. And detailed planning algorithms of basic variables are designed as follows. Extended CoP and CoG trajectories are planned by utilizing sagittal guided strategy and boundary optimization. Ankle point trajectory of swing leg is planned by using the upper command and Extended CoP trajectory. CoC trajectory is worked out by using kinematics between CoG and CoC. According to essence of online motion planning and the difference between online and off-line planning, application of proposed method is discussed and detailed flow combining online planning with real-time control is shown. In order to resist disturbance during humanoid walking, a CoG motion compensation based algorithm is designed. In the end, simulation using "Virtual Blackmann" and walking experiments of "Blackmann" are carried out to validate the feasibility of proposed online motion planning method and effectiveness of proposed disturbance rejection algorithm, respectively.Inverse kinematics (IK) calculaiton is indispensable in humanoid online motion planning and control. In order to check up the feasibility of planning results expressed by the basic variables set and compute the joints'angles accurately online, an integrated algorithm is designed. Proposed IK algorithm synthesizes the geometry based method and damped least squares method after topological configuration analysis of humanoid lower body. Then, simulation is carried out to compare designed IK algorithm with Pseudo inverse method, and the results validate the stability near singularities and feasibility of presented IK algorithm.Detection of force and torque during swing foot-ground contact can help to establish geometric pattern of ground for online motion planning. Based on description of humanoid swing foot landing course, the basic control strategy is determined, and several kinematics and statics related problems are analyzed. Then several representative foot-ground contact modes are transformed into an equivalent single point contact mode, and an active compliant theorem based fuzzy controllor is designed. Combined with online motion planning, application of proposed landing control method in online planning and control is addressed. In the end, simulation results of swing foot landing on slope and "Virtual Blackmann" walking on irregular terrain are shown to validate the effectiveness of proposed method.Finally, the main content of this dissertation is summarized and key points of future research are discussed.