Project Duration:
July 2022 – February 2023
At the renowned Robotic Systems Lab (RSL) I designed and built a jaw gripper for the lower limb of the four-legged robot ANYmal for my final thesis of the Master of Science in Robotics, Systems and Control at ETH Zurich. Traditionally, an additional arm is mounted to the back of four-legged robots for manipulation purposes, which increases their complexity and decreases the available payload. An example for this approach is the Spot Arm of Boston Dynamics. However, a four-legged robot is capable of standing on only three legs, as Philip Arm demonstrated in his paper “Pedipulate: Enabling Manipulation Skills using a Quadruped Robot’s Leg”. Inspired by the feet of primates, we had the idea to mount a gripper to the foot of the robot and use it for manipulation tasks.
Gripper & Wrist
After a lot of consideration, we decided to attach a gripper to the existing shank using an active wrist to transform between walking and grasping. This has the advantage, that we could use the proven foot for walking, and we would only need to focus on designing the gripper and wrist. Additionally, the wrist can orient the gripper for an optimal grasping direction. For my thesis, I focused on the potential task of sampling stones on the moon with an ANYmal.
Jaw Gripper
We developed a jaw gripper, which is backdrivable and has a transparent power transmission. This enables regrasping and grasp detection by estimating the grip force only using the measured motor torque. The gripper is driven by a brushless motor and a ball screw. To synchronize both jaws and prevent jamming of the linear axis, I borrowed an idea of my Robust Hand project by adding rolling contact joints. It has the same functionality as spur gears would have, without the drawbacks of backlash and the need for lubrication. The jaw gripper is very lightweight and adds only one kilogram to the robot’s mass.
Backdrivability
The gripper can easily be backdriven by hand, hence offers a transparent power transmission.
Conclusion
This project marked the first time I designed for manufacturing in metal. Compared to designing for 3D printing in plastic, it required significantly more consideration and planning to get it right. I learned a lot about setting tolerances, designing bearings, and preparing drawings for manufacturing. At the same time, I could use my existing knowledge of ROS and low-level motor controllers to bring the finished project to a level where researchers could easily use the foot gripper.
However, there is still room for improvement, some of which were addressed in the lab. For example, the wrist was not robust, as seen in the video where the gripper shakes a lot while walking, and needed a redesign. Additionally, the estimation of the grip force is not yet accurate and requires further investigation.
