Detecting The Rotor Orientation And Control Strategy Of A Permanent Magnet Spherical Stepper Motor For Humanoid Robot Joint Applications

In recent years, the research of spherical motors has attracted wide attention, because the spherical structure could work at any orientation in three-dimensional (3D) space, especially when used to control the elbow, shoulder joints and arms of the humanoid robots. In addition, spherical motors also play an irreplaceable role in 3D space measurement of intelligent instruments and industrial control in multi-dimensional space where high precision is essential. As a drive organ with many freedoms, spherical motors could simplify the system structure and improve the system performance in steady state, thus becoming a frontier in the current research of mechanic-electrically driven elements. Based on the analysis of the development and trend of motors with many freedoms, this dissertation has made a deep study on a new permanent magnet (PM) spherical motor for humanoid robot joint application.PM spherical stepper motor is a new kind of spherical motor, which is characterized by simple structure, small size, lightweight, high torque and easy control. This paper is based on the PM spherical stepper motor model introduced by Gregory S. Chirikjian et al of American Hopkins University. First, it introduces the physical structure of the motor; then it gives the analyzing result of the 3D air-gap magnetic field and the typical torque characteristic, and meanwhile it analyzes its movement mechanism based on the above results.The paper presents three methods to detect the rotor orientation in 3D space. Due to its particular structure and rotational character, all rotational transforms around the invariable origin constitute into 3-D rotation group, which the paper uses as the basic tool of description. According to random encoder on the surface of the sphere, 96 optoelectronic sensors are used to identify the rotor orientation in order to insure the highest resolution with the finite and discrete sensors. Under this encoder rule, the paper investigates the resolution of the optoelectronic sensors and the result showsthat the resolution could be as much as to≤1~0 . The paper also establishes state space as datastructure to describe the problem, uses graph search method to start single-node search, and establishes evaluation function to start heuristic search, thus greatly reducing the search space greatly and avoiding effectively "combination explosion", thus error lies on the sensor resolution in the current rotor orientation. This method has higher efficiency for the goal node with nearer distance, accordingly suitable for the stepper motors with smaller step angle. In order to improve the search efficiency of remote goal node, however, the paper investigates the global optimization with hybrid genetic algorithm (HGA). To overcome its considerable calculation and iterative course, the paper presents a simple method for global optimization, avoiding a blind search and enormous calculation of random nodes in the origin group. This paper introduces a control method of the PM spherical stepper motor based on table-lookup, calculates the magnet / coil pairs that meet the command of rotation in the current orientation and have them recorded. It also establishes a table that contains all possible rotational trajectory of a certain object located on the surface of the sphere and makes sure that the electrified coils drive the spherical motor along a given trajectory. However, the table needs relatively large memory, which results in the low efficiency of table-calculating and table-lookup. To avoid this, the paper presents a simple of open loop control method of the spherical motor, uses quaternion as the mathematical tool to describe the rotation. The calculation is simple and it is easy to obtain the rotational factors (rotational axial and rotational angle) in several rotation combinations. Thus the paper deduces the mathematical relation between the coil commutation and the rotational factors, and look up the magnet / coil pairs nearest to the given axis. Errors of several rotation combinations are cumulative error. When the error is beyond the angle threshold, orientation correction is needed. Meanwhile, closed loop control is investigated with the feedback signal of rotor orientation. When noise occurs or motor parameter varies, the rotor could be rapidly oriented, and then the magnet/coil pairs that meet the command of rotation are calculated again.