Actuators and Sensors

Biomechatronics 

This research has been motivated by the role of dexterous actuators and sensors for measurement and control of high precision, intelligent machines for robotics and manufacturing automation [1]. Dr. Lee began his activities in the mid1980 with contributions to the development of a three degreesoffreedom (DOF) inparallel wrist mechanism [24] and its micromotion version driven by piezoelectric actuators [5], and an early (1988) proposal for a 3DOF variablereluctance stepper [6] as well as the first showing the important differences between the VRSM and the popular threeconsecutivejoint wrist or spherical wrist motors of a different kind.
Variable Reluctance Spherical Motor (VRSM) Dr. Lee and his students conducted some of the first detailed study on the creation of a three degrees of freedom (DOF) balljointlike variablereluctance spherical motor (VRSM) to provide roll, yaw, and pitch motion in a single joint. This conceptualization of a VRSM, which eliminates constraints imposed by the tradition of threeconsecutiverotationaljoints, has represented a significant breakthrough in new designs of multiaxes directdrive systems that generate smooth, singularityfree, precision motion directly in the target direction in 3D space. This was followed by some of the first detailed studies provide a rational basis for the design [79], torque modeling and control strategies [1012] (based on the concept of distributed, redundant actuation) and noncontact multiDOF encoders for the VRSM, both theory and experimental validation.
Around themid 2000s, Lee and his students have investigated the feasibility of designing a spherical wheel motor (SWM) [13] for applications (such as car wheels, propellers for boats, helicopter or underwater vehicle, gyroscopes, and machine tools) where dexterous orientation control of a continuously rotating shaft. Unlike a VRSM where the stator permanent magnets (PM) and the rotor electromagnets (EM) are placed on locations following the vertices of a regular polygon, equallyspaced magnetic poles are placed on layers of circular planes for a SWM. This enables the shaft to spin using a switching controller while allowing the shaft to incline much like a VRSM. A similar testbed was developed in Singapore [1415] for measuring the magnetic fields of the PMbased rotor and hence realtime computationof the Lorentz torque due to the current flow through the stator coils; validation of these meausred data against the method of magnetic field reconstruction using measured boundary conditions [16].
Unlike a VRSM or a SWM where the rotor PMs are embedded in the “ball,” the PMs of the 3DOF spherical orientation stage are housed in the socketlike rotor assembly. Both forward and inverse torque models are required in the design and control of a 3DOF spherical orientation stage. The forward model simulates the magnetic torque for a given set of electrical inputs. The inverse model (required for both design analysis and realtime control) computes an optimal set of electrical inputs todeliver the specified torque. Two methods commonly used in calculating the forces between stator EMs and rotor PMs of a spherical motor are the Lorentz force equation and the Maxwell stress tensor. These methods require solving the magnetic field and computing a volume or surface integral to derive the force model. As general closedform solutions are not available, the volume or surface integrals are often solved numerically. Tthe interest to optimize the design of taskoriented 3DOF spherical motor has motivated us to develop alternative techniques for analyzing magnetic fields and forces. This effort has led to the method of distributed multipoles (DMP) [17] that computes the 3D magnetic field of a PM in a closed form. We further proposed an improved method to derive an equivalent PM (ePM) such that the magnetic field of the original ML EM can be characterized by DMP. With both the PMs and EMs modeled as DMP, the magnetic forces on the system can be calculated using the Maxwell stress tensor method or the dipole force equation [18]. Unlike other commonly used methods that require to calculate a timeconsuming numerical (surface or volume) integral to derive the force, the dipoleforce equation (based on distributed dipoles replacing surface or volume integrals by summations) offers closedform solutions dramatically reducing computation time by as much as four orders and has been validated against known analytical solutions and experimental results.
Sensors for MultiDOF Spherical Motor and Orientation Stage
In the early 1990s, Dr. Lee coinvented an imagereading system (see Machine Vision), and with Parker [19] contributed to the vision system modeling and methodology for designing efficient vision systems that might otherwise require timeconsuming and expensive physical experimentation. The imagereading system has provided a basis for developing a 3DOF visionbased orientation sensor [2021] that reads encoded gridlines printed on a spherical surface. Lee and Zhou [22] extended the visionbased method but used microscopic surface features (instead of dedicated pattern) to determine motion changes, which led to the design concept and theory of a dualsensor system capable of measuring 3DOF planar and spherical motions in real time. The concept feasibility of two prototype 3DOF dualsensor systems for measuring the instantaneous center of rotation and the angular displacement of a moving surface was demonstrated experimentally.
More recently, a novel method that determines the displacement and force/torque of a PMbased electromagnetic actuator using existing magnetic fields with lowcost Halleffect sensors has been developed in [23]. Using a linear voice coil motor (VCM) as a platform for analysis, the experimental performance of a 9sensor fieldbased system was compared against an optical encoder [24]. With a sinusoidal reference signal (0.2Hz frequency, 4mm amplitude), the corresponding positional estimate provided by the HallEffect sensors under PID control for one time period (shown below). The encoder controlled response follows a ‘staircase function’ (each step corresponds to the encoder resolution of 42.33μm) and is unable to adequately reproduce the peak and trough of the sinusoidal reference command. The fieldbased system, which by virtue of the analog outputs of the HallEffect sensors and continuous model of the ANN, permits interpolation between discrete training set points. This result in an overall smoother response as the fieldbased system is able to interpolate multiple data points within the encoder resolution as illustrated by the insets. In addition, while the actual peak and trough of the response can be inferred from the HallEffect sensors, the discrete nature of the encoder prevents a similar extraction. The demonstrated findings offer potentials in several other unique applications where magnetic fields play a role.


Professor KokMeng Lee
