ABSTRACT
The idea of sensorizing a strain wave gear to measure the transmitted torque has been reported since the 1980s. The strain in the elastic flex spline is typically measured by strain gages attached to it. The resulting voltages relate to the transmitted torque in the gear. However, periodic inaccuracies in the measured torque signal (sensing ripple), resulting from positioning inaccuracies of strain gages on the flex spline, prevented this technology from being used outside a lab environment. Regardless of these difficulties, measuring the torque directly in the strain wave gear would bring many advantages, especially in robotic applications, where design space is highly limited. Traditionally, robotic joints are equipped with link-sided torque sensors, which reduce the available design volume, lower the joint stiffness, and require complex cable routing. This paper presents an experimental study of a novel sensorized strain wave gear named RT1-T, which was developed by Schaeffler Technologies. The study was implemented on a joint testbed, including a high-resolution reference torque sensor at the link side. In addition to the measurement accuracy and linearity, a torque ripple analysis is performed. The joint torque control capabilities are determined along dynamic trajectories and compared to the performance achieved with a link-sided reference sensor. The sensor employed in the testbed has a static torque error of 0.42 Nm and an average closed-loop torque control error of 0.65 Nm above the reference sensor.
ABSTRACT
We propose a fault-tolerant estimation technique for the six-DoF pose of a tendon-driven continuum mechanisms using machine learning. In contrast to previous estimation techniques, no deformation model is required, and the pose prediction is rather performed with polynomial regression. As only a few datapoints are required for the regression, several estimators are trained with structured occlusions of the available sensor information, and clustered into ensembles based on the available sensors. By computing the variance of one ensemble, the uncertainty in the prediction is monitored and, if the variance is above a threshold, sensor loss is detected and handled. Experiments on the humanoid neck of the DLR robot DAVID, demonstrate that the accuracy of the predicted pose is significantly improved, and a reliable prediction can still be performed using only 3 out of 8 sensors.
ABSTRACT
In waterjet surgery, a thin high-pressure jet is used for dissections and surface abrasion of soft tissue. This selective preparation method preserves nerves and vessels, whereas the surrounding soft tissue is washed away. OBJECTIVE: The aim of this study is to enhance the application field of this technique by resolving technological limitations. METHODS: A technical task definition of handling a hand-guided waterjet applicator is derived from the literature. All reported procedures require to follow a trajectory superimposed with an oscillating movement. By introducing a robotic system and a specialized kinematic approach, the limited dexterity of the waterjet applicator is increased. Additionally, the system provides assistance by automatically performing parts of the task. RESULTS: The method is applied to two different procedures: a minimally invasive dissection and a surface abrasion for open medical treatments. On the basis of experiments with gelatine phantoms, the performance of the method is shown for both procedures. CONCLUSION: In the minimally invasive use case, the reachability limited by the conventional manual tools is extended by the capabilities of the robotic system. Simultaneously, the handling is simplified by automation of the superimposed oscillation. In the surface abrasion case, a dense coverage of the treated area is achievable. The risk of cross infections could be reduced by spatial separation of patient and staff. SIGNIFICANCE: Thus, the waterjet technology can be fully integrated into robotic surgery systems and benefit from their inherent abilities.