HI, I'M DAVID
A controls engineer with a passion for robots.
Berkeley M.Eng M.E. '20
UVA B.S. M.E. '17
ABOUT
I am a Robotics Engineer and a recent Master's graduate of the University of California Berkeley where I studied Mechanical Engineering concentrating in Control of Robotic and Autonomous Systems. Currently I work for Dexterity, a robotics startup developing robots for logistics and supply chain with human-like dexterity. Previously, I worked for Rolls-Royce for two years in a rotational program including positions in Control Systems Engineering, Design Engineering, Digital Manufacturing Engineering, and Manufacturing Engineering. In 2017, I graduated from the University of Virginia earning my Bachelor of Science in Mechanical Engineering.
​
I'm passionate about intelligent technology - technology which is aware of and responds to the world around it. In the future, I want to work with a dynamic team on cutting edge technology to push the future of "the possible". Nevertheless, I'm a strong proponent of the Jeffersonian Engineer - an engineer who understands the ethical and societal impacts of technology.
​
Outside of work, I am an avid backpacker. Summer of 2019, I walked 550 miles across Spain on the Camino de Santiago - experiencing the local culture and making friends from around the world along the way.
PORTFOLIO
CONTROL OF AN AUTONOMOUS UNMANNED AERIAL VEHICLE (UAV)
1st Place Among 75 Students in Project Competition
After many tests, including one where I had to duck out of the way of the wild UAV as it sped at me, crashed, and broke one of the arms, my team and I developed a controller permitting autonomous navigation and control of our UAV flying over a barrier and landing on a specific spot on the floor. The controller was written in C++ and was designed using fourth order cascaded control to control position, velocity, orientation, and angular velocity. Moreover, we included a low gain PI controller in loop for greater landing precision. After characterizing the motors and sensors, we also designed a state estimator using the on board gyroscope, accelerometer, height sensor, and optical flow sensor correcting for bias and drift.
Toward Adapting Humanoid Robots to Aid First
Responders
Master's Capstone Project
Fung Institute Mission Award Winner
To reduce the risk first responders face, our project took the first steps toward making legged robots fast and mobile enough to take their place. This is possible because we enabled a bipedal robot, Cassie Cal, to transition from walking to stepping onto a static platform — the foundation needed for Cassie to step onto a dynamic platform, such as a Segway or Hovershoes.
​
To do so, we solved a nonlinear constrained optimization problem offline to formulate a gait library. These gaits were then implemented via virtual constraints approximately zeroed with PD control. Finally we developed a finite state machine to control behavior.
​
In this project we had three primary novel contributions. First we addressed underactuation and static instability of line-footed robots via designing a strategy for Cassie to dynamically step on a platform. Second, we refined the joint controller to enforce asymptotic stability within one step. Third, we encoded event transitions into the trajectory optimization to precisely control the COM.
Many thanks to our advisor, Professor Koushil Sreenath, our graduate mentor, Bike Zhang, and the entire Hybrid Robotics Lab at UC Berkeley.
NONLINEAR MODEL PREDICTIVE CONTROL OF A ROBOT RIDING HOVERSHOES
Autonomous Motion Planning
This project presents a nonlinear Model Predictive Control formulation for trajectory tracking of the bipedal robot, Cassie, mounted on Hovershoes. The MPC utilizes a modified bicycle kinematic model and a reduced form of the full 62 state vector. A convex optimization problem was formulated for trajectory tracking under system constraints. My team showed in this project through simulation that this method can successfully track a one-period sinusoidal wave trajectory as well as a trajectory around a three meter box obstacle.
CONTACT FORCE OPTIMIZATION CONTROL OF A BIPEDAL ROBOT
Robotic Balancing Controller
My team and I developed and implemented, in simulation, a balancing controller for Cassie, a bipedal robot with 20 degrees of freedom. Our controller utilized a contact force optimization control strategy which first translates a desired center of mass position into a desired wrench with a PD controller, second computes the optimal force to apply at the foot contact points to achieve the desired wrench, and third maps that force to joint torques using the Jacobian. The optimization was solved with a quadratic cost function weighting the desired force greater than the desired moment and constraining the force to lie within a linearized friction cone. This simulation shows the robust nature of the balancing controller as Cassie is hit with a perturbation force but does not fall over.
MODEL PREDICTIVE CONTROL FOR URBAN AIR MOBILITY
Optimal Control
Using Model Predictive Control (MPC), an Unmanned Arial Vehicle (UAV) was simulated flying in 3D space between set destinations. The Constrained Finite Time Optimal Control problem (CFTOC) is weighted to prioritize optimization of input power, velocity and deviation from the reference trajectory. Furthermore, the UAV is constrained to fly within an admissible tube, limiting the UAV’s maximum displacement from the reference trajectory. This simulation shows the N step open loop prediction of the CFTOC in cyan, the closed loop trajectory over time in dark blue, and the reference waypoints in red.
GET IN TOUCH
