Study On Motion Planning For Omni-directional Mobile Humanoid Robot

Omni-directional mobile humanoid robots have advantages for their agile movingability on planar floors and the ability to imitate the operation ways of the human’s arm.As a front research topic in service robots field, they have a wide application prospect inhousehold management services, welcome and guide the guests in restaurant or any othercommercial places.Because the work environment for omni-directional mobile humanoid robots is usuallydynamic and unstructured, robots must have the ability to adapt variable environment.Furthermore as omni-directional mobile humanoid robots have complicated structures andmany degrees of freedom, it is a great challenge for on-line motion planning with collisionavoidance when the geometry constraints, kinematic constraints and driving forceconstraints are taken into account.In order to study on on-line motion planning for omni-directional mobile humanoidrobots transporting materials in dynamic and unstructured environment, a robot prototypehas been developed. This dissertation concentrates on the motion whose position andposture in the initial and termination are restricted, as well as the motion whose posture isrestricted from the initial to the termination. Some key problems including singularityavoidance, collision-free path planning, time-optimum trajectory planning and robustnessof control are studied. Simulations and experiments have also been performed todemonstrate the methods presented in the dissertation. The mainly research work are listedas follows.The inverse kinematics calculation and singular problems are discussed. Because thecontinuity of the posture transform in work space is necessary, the quaternion is used todescribe its rotation. As the omni-directional mobile humanoid robot can plan its motionautomatically, we apply a method of changing the trajectory of end-effector in work spaceto avoid its singularity. A quadratic optimization function including multi-objectives isconstructed for optimal motion calculation. It is necessary for the omni-directional mobile humanoid robot on-line determines itsmotion without collision occurring in dynamic and unstructured environment. As the robothas many degrees of freedom, a path-planning method synthesizing probabilistic roadmapmethod (PRM) and rapidly-exploring random trees method (RRT) is proposed which takesthe completeness and efficiency of algorithm into consideration. The probabilisticroadmap method provides roots of RRT, and search the shortest collision-free path in thejoint space. While the RRT search the joint space nodes meeting the pose constraintsrequirement in the work space. This proposed method can plan collision-free paths for theomni-directional mobile humanoid robot by sensing the3D models in the environment.The path should be optimized to ensure physical sensors work reliably as the sensorsrange is usually limited.In this paper, it is proposed a time optimal trajectory planning method based on thelimited capacity of the robot motor drive and the discrete control model. This method usescubic B-spline curve fitting the discrete path nodes, and the time optimal robot trajectoryis obtained by constructing a discrete trajectory net. Meanwhile, the time optimaltrajectory is discretized to control real robot, and the robustness of trajectory trackingcontrol is demonstrated when the disturbances exist in practical using.In order to demonstrate the effectiveness of the proposed method, simulation softwareand an omni-directional mobile humanoid robot are developed. The details of the motiongeneration and motion implementation system are introduced. The whole system providesthe prerequisite and foundation of experimental verification for the proposed motionplanning algorithms. Furthermore it is one of the key technologies for developing theomni-directional mobile humanoid robot.The on-line motion planning research for the omni-directional mobile humanoid robotwill help to improve its intelligent of motion, so as to enhance its adaptability in practicalapplication. Both in theory and application the research in this dissertation has somereference value.