Masashi Horiuchi

Accuracy improvement of built-in torque sensing for Harmonic Drives

Built-in torque sensing for Harmonic Drives is attractive since it maintains mechanical characteristics of the gear while providing detection of the transmitted torque. Torque sensing by using strain gauges has been studied, but is not widely used yet due to a relatively high signal fluctuation, which is generated by the gear operation. Characteristics of the signal fluctuation are analyzed in this paper, and a method to effectively compensate the signal fluctuation is proposed. The signal fluctuation is perfectly compensated by adjustment of the strain gauge sensitivities. A minimum number of strain gauges needed to compensate the signal fluctuation is derived. The experimental result with three strain gauges compensating the basic frequency component of the signal fluctuation is shown.

Performance of gain-tuned harmonic drive torque sensor under load and speed conditions

This paper addresses the topic of torque sensing by using strain gauges, which are cemented directly onto the flexspline of a Harmonic Drive gear reducer. Conventionally, two or four strain gauges are used to reduce the ripple signal, which is generated by the gear operation. However, our analysis shows that an odd number of strain gauges is more suitable for the ripple compensation. The method does not need online calculations and reduces needed accuracy of the strain gauges positioning, so it can be practically implemented. Two techniques to define the gains are presented in this paper. One is based on a mathematical model of the fluctuation signal, and the other employs a heuristic approach. Practical results show effectiveness and usefulness of the proposed method. In the paper, we present the new method and evaluate its performance under load torque and rotational speed conditions. No significant deterioration of the performance under load torque and at various rotational speeds was confirmed.

Velocity dependence of the characteristics of harmonic drive built-in torque sensing

We have proposed practical torque sensing which utilizes a flexible part of a harmonic drive gear. The sensing technique provides joint torque sensing without reducing stiffness of the robot and changing the mechanical structure of the joints. The characteristics of the torque sensing have been studied under an immovable condition. The dependence of characteristics on the rotational velocity have not been discussed. We describe the characteristics under rotational conditions of the harmonic drive. The experimental results show that the accuracy of the torque sensing under high velocity rotation is 2% of the gear torque capacity.

Improved performance of built-in torque sensing for harmonic drives

Torque sensing built into harmonic drives has advantages in its size and stiffness comparing to the torque sensors added to the robot joints. A relatively high fluctuation signal of a built-in torque sensor is a drawback in its practical applications. However, we present a simple and effective method to compensate the fluctuation signal and thus improve an overall accuracy of the harmonic drive built-in torque sensing. The method does not require precise positioning of the strain gauges, does not require online-estimations of the fluctuation signal, and does not increase the number of applied strain gauges. It requires initial tuning of the gains. For the signals of the separate strain gauges, two techniques to define the gains are presented: one is based on a mathematical model of the fluctuation signal, and the other employs a heuristic approach. Practical results show the effectiveness and usefulness of the proposed method.

Ripple compensation of harmonic drive built-in torque sensing

Built-in torque sensing in Harmonic Drives enables torque sensing without assembling additional torque sensors into mechanisms where Harmonic Drives are already present. The sensing principle has been known for about ten years, but it is not widely utilized yet, mainly because of a relatively high ripple signal in the sensing output, generated by the gear operation. Ripple is difficult to be compensated due to inaccuracies of the strain gages positioning and in geometrical properties of the gear and its assembly. Increased number of applied strain gages reduces the ripple, but does not eliminate it. In this paper, the authors present a new method to effectively compensate the ripple, it is based on a periodic characteristic of the ripple signal itself and proposes use of separate amplifiers for each of the signals from the strain gages. Gains of the amplifiers are tuned so that the ripple signal is compensated. A mathematical model of the ripple signal, and a method to calculate the tuned gains is studied. Minimum number of strain gages needed to compensate the ripple signal is derived. The method is successfully confirmed by experiments.