Torque Sensor Embedded Actuator Module for Robotic Applications
ABSTRACT
Measuring accurate torque is a core ability of advanced robotic technologies. However, embedding torque-sensing systems into robot joints is still a challenge due to limitations including size, weight, and cost. Moreover, the influence from axial forces/torques (x- and y-axis) and the sinusoidal effect from the harmonic drive make it difficult to measure torque accurately. This paper presents a novel actuator module for robotic applications that includes a torque sensor, harmonic drive, motor, and encoder. The torque sensor adopts the capacitive sensing scheme, which can allow for self-decoupling of the axial influences through the use of a symmetric arrangement of sensing cells. Moreover, the torque ripple cancellation method, which can reduce the sinusoidal effect from the harmonic drive, is introduced. Finally, the detailed design of the actuator module is described, and its performance is experimentally evaluated.
EXISTING SYSTEM :
ROBOTS are extensively used in various fields as service robots, surgical robots, rehabilitation robots, etc. Above all, applications involving human-robot interaction have rapidly increased in recent years. However, caution is still applied when robots are operated around humans. In addition, in the industrial robotics field, there are restrictions regarding the allowed operating range around humans. To overcome this problem, the capability of delivering accurate force/torque information and detecting collision information are the key requirements for extensive robotic systems. Furthermore, measuring accurate force/torque can provide a solution for current robotic systems in which the motor current is used to estimate the operating torque and to detect collisions.
The motor current cannot provide accurate torque information because of the unpredictable disturbance from gear transmissions and assembly errors. Moreover, current-sensing-based force control requires considerable data analyses and complex data processing tasks . In response to these problems, attaching torque sensors to each robot joint has been proposed by many research groups. To date, many research groups have developed torque sensors based on strain gauge. Kuroki et al. studied the material of the strain gauges to obtain high sensitivity without using expensive signal amplifiers . They found CrN-STFs (Nitrogen added Chromium alloy thin films) that can provide the high sensitivity on highly stiff sensors. However, owing to the difficulty of attaching a number of strain gauges to the flexure hinge, it has limitation on the mass producton. Lee et al. studied the structure of the flexure hinge to produce identical deformations at every point.
As this resolved the difficulty in attaching strain gauges, this can provide a solution for the difficult assembly process. However, a signal amplifier is required to convert the resistance to torque. Some groups have also proposed an optical-type torque sensor based on optical encoders, inductive transducers and photo detectors. However, the use of this type of torque sensor requires a transducer. In addition, there are several critical problems involved in embedding the torque sensor in the robot joint.
The first problem is the influence of axial forces/torques (x and y-axis). While operating the robot manipulator, axial forces/torques are applied to the joints due to the weight of the manipulator and its load. As a result, the measurement of accurate torque is disturbed. In general, bearings are used to counteract these axial influences, but they increase the mass of the robot. The second problem is the sinusoidal influences from the harmonic drive. The harmonic drive generates a high gear reduction ratio using an elliptically shaped wave generator and a flexible flexspline. While this operation generates the elliptical deformation of the flexspline, the directly linked torque sensor detects the unwanted torque information referred to as “torque ripple”. This causes a significant interruption for accurate torque measurement.
PROPOSED SYSTEM :
In this paper, a torque sensor embedded novel actuator module was introduced. The torque sensing and decoupling ability of the axial forces/torques were evaluated using four symmetrically located sensing electrodes. To increase the sensitivity of the sensor, an advanced sensing method was adopted using the orthogonal arrangements of two sensing electrodes. The sensor was realized through the theoretical analysis of the deformable plate. As a result, the proposed actuator module was composed of four housing parts, two input shafts, a torque sensor, a harmonic drive, a cross-roller bearing, an encoder, and a motor. Its assembly process only required several bolting connections. The performance of the developed actuator was compared with commercialized torque sensors and actuators. It was found that the torque-sensing ability of commercialized torque sensor such as TFF350 and Mini85 showed better sampling frequency and high resolution. However, these sensors are based on attaching the strain gauge to the sensor, and thus, they have problem with the mass production. Furthermore, their performance is largely dependent on the signal amplifier. In addition, the developed module showed larger sensing range and higher sensing frequency than ANYdrive and K-75. Embedding torque-sensing systems into robot joints is still a challenge due to limitations including size, weight, and cost.
CONCLUSIONS
In this paper, a torque sensor embedded novel actuator module was introduced. The torque sensing and decoupling ability of the axial forces/torques were evaluated using four symmetrically located sensing electrodes. To increase the sensitivity of the sensor, an advanced sensing method was adopted using the orthogonal arrangements of two sensing electrodes. The sensor was realized through the theoretical analysis of the deformable plate. As a result, the proposed actuator module was composed of four housing parts, two input shafts, a torque sensor, a harmonic drive, a cross-roller bearing, an encoder, and a motor. Its assembly process only required several bolting connections. The performance of the developed actuator was compared with commercialized torque sensors and actuators. It was found that the torque-sensing ability of commercialized torque
sensor such as TFF350 and Mini85 showed better sampling frequency and high resolution. However, these sensors are based on attaching the strain gauge to the sensor, and thus, they have problem with the mass production. Furthermore, their performance is largely dependent on the signal amplifier. In addition, the developed module showed larger sensing range and higher sensing frequency than ANYdrive and K-75. In this study, only single actuator is used to evaluate the decoupling ability of the axial influences and the torque ripple cancellation. As an extension of this study the evaluation in a multi-DOF(degree of freedom) system will be conducted in the future.