.jpg){kind=small}
Projects
My research focuses on the design of state-of-the-art robotic leg hardware. Specifically, I design robotic knees, ankles, and feet to address the challenges of bipedal locomotion in unstructured environments, both on humanoid robots and disabled humans. My expertise in actuators and force sensors combined with my experience with force and impedance control enables me to collaborate well on whole-system integration of highly dynamic robots. While these projects demonstrate specific skills I've acquired, the breadth of this skillset speaks more generally to my ability to adapt to challenges and master new skills needed to complete new projects.
Tesla Optimus Ankle Actuators
As the owner of the ankle actuators on Optimus, I integrate the design of these linear actuators with multiple sub-systems, including motors, gears, sensors, linkages, motor controllers, bearings, harnesses, and the surrounding structures. Due to the reliance of the ankle actuators on their series and parallel structures, I also collaborate the development of the tibia, foot, and prismatic differential linkage. These duties, in addition to my biomechanics knowledge and experience in powered prosthetic knee design, give me a broad spectrum of expertise in robotic leg design. Regarding robotic gearing, in addition to my experience with standard and inverted roller screws, my position on the actuators team, and specifically with torque sensor design, exposes me to the design tradeoffs and integration challenges of strain wave, cycloid, and worm gearings.
Tesla Optimus Force/Torque Sensors
As the lead designer of all force and torque sensors on Optimus, I design structurally-integrated force sensors for all linear actuators and torque sensors for all rotary actuators, as well as 3-DoF ankle and 6-DoF wrist force/torque sensors. I lead a team of four design engineers and two interns to design and calibrate all sensors, as well as perform validation tests to ensure the sensors have industry-leading dynamic range and stability and are well-protected from cross-talk, thermal effects, EMI, overloading, and internally-generated noise (e.g. ovalization and eccentricity of a strain wave gear). Through rigorous design and testing, Optimus actuators have high-resolution (<0.1% full-scale), high-bandwidth (>10 kHz) force and torque sensors with <1% full-scale error from the combined effects of non-linearity, hysteresis, non-repeatability, cross-talk, thermal effects, and EMI.
​
Variable Transmission Knee Prosthesis
To address the deficiencies of state-of-the-art passive knee prosthesis, several researchers have developed powered knee prostheses. While these devices restore positive mechanical power, their large motors and/or transmissions increase actuator impedance. This reduces the ability of the user to freely swing the prosthesis in a biomimetic manner, which makes the swing motion of powered knees deficient relative to passive knees. To decrease the joint impedance for free swinging behavior, this work developed a novel variable transmission for a knee prosthesis, which adds powered functionality without sacrificing passive behaviors necessary for restoration of normal knee function. This work was recently accepted for publication in IEEE Transactions.
Unified Passive Walking Controller
_JPG.jpg)
State-of-the-art commercial passive knee prostheses are capable of biomimetic function during level-ground, up-slope, down-slope, down-stairs, and backwards walking, as well as when sitting down. However, none of these control systems have been fully-disclosed, and no publication has described a comparable control system. This work developed a unified passive control system for a variable transmission knee prosthesis with similar performance to a commercial prosthesis in all the aforementioned activities. Resistive behaviors were achieved with motor braking, which regenerates energy and reduces power requirements for walking.
Powered Knee Assistance Controller
_JPG.jpg)
The design of a powered prosthetic knee capable of free-swinging behavior and the development of a passive control system with performance similar to a commercial passive prosthesis opened the door to a new approach to powered knee control: powered assistance. Rather than use power to always control the motion of the prosthetic knee, this approach assumes that power generation should only be used for 1) activities requiring net knee extension during stance-phase, such as stair ascent and sit-to-stand; and 2) activities requiring powered swing, such as the swing-phase during stair ascent. When not providing power, the prosthesis should remain strictly-passive. In other words, the device should be passive when possible, and powered only when necessary. This approach has high electrical efficiency and preserves user agency. This work won the best paper award at ICORR 2022.
Powered Ankle Stair Controller
Stairs are commonplace in society but can be difficult obstacles for individuals with below knee amputations. This work focused on developing a control system for a powered prosthetic ankle to ascend and descend stairs. Additionally, a method for volitionally switching between stair ascent, stair descent, standing, and level ground walking was also developed based solely on user motion. The control approach was able to restore many features representative of unimpaired gait to the user relative to his passive daily-use prosthesis, including shock absorption during stair descent and net positive power delivery during stair ascent. Results of this experiment are published in IEEE Transactions.
Strain Gage Based Load Cell
This custom Maltese cross strain-gage based load cell is able to measure axial loads up to 1200 N in the presence of sagittal, frontal, and torsional moments of 130 N·m, 20 N·m, and 20 N·m, respectively, which are common peak loading profiles for a knee prosthesis with 100kg user. The load cell has a low profile and a mass less than 100g. The load cell was tested with axial loads coupled to varying moment magnitudes. Output from the Wheatstone bridge was linear with slight hysteresis for all loading conditions, with less than 20% cross-talk for the maximum loading and combined moment conditions. This interference did not affect load sensing for control of a knee prosthesis. This design has been purchased by the Parker Hannifin corporation.
Digital Grip Gauge
Lockheed Martin Manufacturing Technologies tasked our senior design team with a project to design, manufacture, and test a digital grip gauge smart tool capable of measuring the grip length of countersunk holes on the F35 aircraft. Using an ergonomic body design, an inductive digital scale, and a novel actuated probe, we designed, manufactured, and tested the smart tool, achieving all design objectives, and producing a tool capable of measuring grip length and transmitting the measurement at twice the desired rate and with an accuracy of +/- 0.001 inch. The digital grip gauge was patented by Lockheed Martin and was estimated to have saved $20 million for the F35 program by improving manufacturing efficiency over the previous manual methods.
Topology Optimized Bracket
This work was completed during an internship at the Center for Limb Loss and Mobility (CLiMB), in support of a project to evaluate the sagittal and coronal stiffness profiles of commercial carbon fiber prosthetic feet during physiological gait using a six-axis robot. The bracket needed to attach a load cell at a 90 degree angle from the robot's output, with a mass constraint based on the robot's maximum payload. Solidworks Topology Optimization was used to inform us how to best distribute mass, and the optimized results were interpreted into a manufacturable bracket design. The final design increased bracket stiffness by a factor of 18 compared to the previously-used design.
Touch Sensing Handrails
This work was completed during an internship at the Center for Limb Loss and Mobility (CLiMB), in support of a project to assess stability of below-knee amputees while walking on level-ground, cross-sloped, and uneven terrain treadmills. Handrails are included on the treadmills for safety of study participants, but they interfere with stability margin estimates from the motion capture system when participants use them. To discriminate when stability margin estimates are valid, I designed and manufactured a set of touch sensors for the handrails that send a digital high/low signal to the motion capture system when the handrails have been touched.
Xbox Kinect 3-D Surface Scanner
This work examined the function, accuracy, and ease of use of an XBOX Kinect paired with Skanect (Occipital) and MeshLab software packages as a low cost solution to 3-D surface scanning and processing. The Kinect was able to accurately model the recorded point cloud as a continuous 3D surface that matched the contour and scale of the test subject surface (shoe insoles). Both Skanect and MeshLab effectively interpolated the smoothing of the 3D surfaces and provided higher resolution imaging than both the unaltered image and the resolution of the 3D printers used in this experiment. 3D printing of the model produced a duplication of the test surface. This work won a best poster award.
