Jumping Robotic Leg

October 2023 - April 2024

TLDR:

• Led the development of motion planning for a jumping cat robot, programming ROS2-based motion sequences for jumping movement, allowing for iterative jump tests and a final 6-inch jump height of a prototype leg

• Conducted torque experiments and a root-cause analysis to troubleshoot and optimize performance of BLDC motors, increasing output torque by 10% during testing phases

• Designed a 60:1 worm gear reduction system to apply high torque to a torsional spring in the leg’s knee with OnShape

This was a project with the Rice Robotics Club. There aren’t any quadruped robots that optimize jumping height (that we could find at least), which led us to take inspiration from a jumping cat and make one. This project lasted for my entire junior year.

I took a focus on the actuation and control of the robot, helping research different motors, gear reduction mechanisms, and energy storage methods for a pouncing movement. We decided on a design that used a 2 degree of freedom leg, with both joints actuated by brushless DC motors.

Additionally, we decided that a spring should be attached to the knee joint, allowing for a charged pouncing motion. The leg would lay prone, and a servo would extend to lock it in place as the motors charge the spring. The leg would then pounce when the servo releases the lock.

Gearbox:

I designed the gearboxes for the robot, utilizing timing belts to achieve a 1:30 gear reduction. We chose to use timing belts because of their simplicity and low cost. However, as the project progressed, we noticed problems with the belts. We had underestimated the torques that they would undergo and the belts would constantly skip. Additionally, we found them to be difficult to assemble with the 3D-printed leg body and found their large size to be cumbersome. We decided to switch all gearboxes to worm gear systems due to their compactness and high gear reductions. Following this I edited the gearbox design to include a 1:60 worm gear reduction, including beefing up the gearbox to account for the higher stresses induced by the gears.

Our final prototype

Motors:

I took the task of readying the motors. We bought brushless DC motors (commonly found on drones) and controlled them with a separate motor controller board with an encoding magnet. The encoding magnet needed to be exactly 1 mm away from a specific chip on the control board, so I 3D-printed a simple holding stand to test the motors that held everything in the right place.

Initially, the motors exhibited strange behaviors, including excessive vibration, low speed and torque, and occasional inoperability. Throughout the next few months I troubleshooted the various issues with the motor using a root-cause analysis. This included perfecting the testing stand to further stabilize the motor and control board, removing a bearing between the motor shaft and the encoding magnet as it was interfering with the magnetic field, and insulating screws between the motor and the control board as they were shorting the board. By March, the motors were running reliably at the torque and speed we desired.

Motion:

A Raspberry Pi was connected to the control boards via CAN bus to allow us to program the motor movements. Under the ROS2 framework, I programmed a jumping script in Python to use for jump tests when the robot leg was fully assembled. The script goes through phases in a jump, beginning with the leg laying prone and being locked into place with the servo. Then the spring in the knee is fully charged, readying the leg. When the servo unlocks the leg, the spring then causes the leg to pounce. In midair, the leg adjusts itself to brace for a landing. During the landing, the forces are absorbed by the spring and the leg returns to its initial position.

Integration:

With the motors working as desired, the integration of the motors to the robotic leg itself began. A similar design to my motor test stand was used to attach the motors and boards to the leg. Once the power-transmitting shafts were connected to the motor, we began testing the individual joints. The legs ran without issue at maximum speeds, but at maximum torque the gears skipped and the power-transmitting shaft on the knee joint kept bending and shifting out of place due to the high axial and radial loads. Since we were very close to our deadline, we had to jerry-rig solutions to fix these issues, including attaching a plate perpendicular to the end of the shaft to keep it from shifting out of the leg. 

We then set up the full leg assembly for jump testing. We encountered more petty mechanical problems. We were only able to get our first successful jump of the leg right before our presentation at the Rice Engineering Design Showcase. Unfortunately, the hip joint didn’t work properly and the jump came exclusively from the spring’s stored energy. Nonetheless, we achieved a 6 inch high jump.

Reflection:

I was excited to work further on optimizing the leg’s jump, but it was disappointing to run out of time. However, it was a challenging project and I was still very proud of our result. I learned a lot, including the intricacies of mechanical design and the complexity of integrating everything together. Working on this project flourished a passion for robotics and mechatronics.

Additionally, I learned about project management. I was promoted to a leadership role in the club and was involved in recruiting and management of the project. We were good at recruiting, mainly because our project was cool and flashy and we tried to lock down members before they were snatched by other clubs. Management could have gone better, as we consistently had overly optimistic plans for the development of the robot. It wasn’t until a month before the deadline that we conceded that we wouldn’t have the time to fully integrate four legs for a fully complete robot.

This project was continued in the next year, focusing on investigating alternate leg designs. However, I (sadly) decided to step down from my position to focus fully on my senior capstone project. I still keep up with their progress and help out a little from time to time, though.