Academic

Full-stack engineering on course-wide robotics competition robot, including mechanical design, electrical schematics, wiring and soldering with custom PCBs, firmware development, and overall integration, testing, and debugging. Overall course-wide undefeated tournament winners!

Developed C code and electronics for PIC32 microcontroller to control DC motors, quadrature encoders, servo motors, optical reflectance sensors, and IR LEDs and phototransistors. Software includes digital and analog I/O, signal processing and filtering, PWM, PID control, and more. All custom PCBs laid out and soldered, software developed with complex hierarchical state machines enabling fully autonomous gameplay.

Designed and manufactured robust chassis, drivetrain, superstructure, and custom belt-driven compliant wheel intake system with 3D printing, laser cutting, and various OTS components. Multiple iteration and testing cycles on intake system for optimized gameplay routines in conjunction with software development.

In collaboration with Andrew Wang and Annika Yong. Linked here is the project website.

Autonomous motor racing using lap-time optimization, intention inference, and custom safe model predictive controller against adversarial agents in simulated environment. 

Implemented an offline trajectory optimization method to determine optimal racing lines on known track(s) with dynamic feasibility, solving minimum lap-time optimization problem in IPOPT with Pyomo DAE solver.

Utilized Gaussian process regression to map between ego and adversarial agent states and local track information using scikit-learn's GaussianProcessRegressor class with Matern kernel to interpolate into predicted future adversary states.

 Integrated optimized racing line trajectories and intention inference adversarial states from Gaussian regression into receding horizon MPC with IPOPT with CaSaDi package, penalizing racing line devation, jerky steering, and intrusion into preset safety distance barrier.

In collaboration with Andrew Wang. Linked here is the project paper, slides, presentation, and experiment videos.

Full-stack engineering on mechatronics companion robot, including mechanical design, electrical schematics, wiring and soldering, firmware development, and overall integration, testing, and debugging.

Developed C code and electronics for PIC32 microcontroller to control LED dot-matrix display, servo motors, IR proximity sensors, capacitive touch sensors, LED arcade buttons, and a 2-axis joystick, including digital and analog I/O and PWM.

Designed and manufactured enclosure and auxiliary chocolate dispenser with laser cut and 3D printed parts and integrated with electronics components for ease of testing and debugging.

In collaboration with Wanyu Weng and Kanishk Chaudhary. Linked here is the project website and code repository.

Developed state estimator and extended Kalman filter (EKF) to integrate into model predictive controller for Indy Autonomous Challenge Dallara AV-21 racecar. 

Improved upon previous state estimator by increasing from 8 to 14 state variables, notably incorporating side-slip angle and slip ratio into calculations with Dugoff brush tire model for higher accuracy state estimation at higher vehicle speeds.

Implemented autograd Python library for vehicle dynamic model and Jacobian calculation with modified EKF incorporating inputs, measurements, and new states into previous 

In collaboration with Andrew Wang, Blake Werner, and Kai Yun. Linked here are the final paper and presentation.

Developed custom dart launcher with vision-based circular target acquisition and path planning with PID control on Sawyer Robot. Designed and manufactured 3D printed dart launcher with spring propulsion. Iteratively improved design with empirical testing with individual launcher module and with robot assembly.

Created target acquisition system using vision tracking to find circular dartboard location, processed with thresholding and Hough transform, and calculated coordinate frame transformations to robot to determine relative pose. Optimal launch trajectory is then calculated with projectile motion equations to determine the desired robot end effector pose for launch.

Utilized ROS and MoveIt (notably MoveIt Commander) path planning with PID control to move robot end effector to desired pose and launch dart with no additional sources of actuation or powered mechanisms. Path planner plans a large set of possible planned paths to find shortest path.

In collaboration with Andrew Wang, Blake Werner, and Kai Yun. Linked here are the project website and a launch video.

Used Matlab script to generate training data, with PDE toolbox to run FEA simulations on 2D grids of randomly placed fibers with set boundary conditions.

Implemented  modified gradient descent algorithm in Python based on weighted combination of fiber adjacency and stress differentials to iteratively generate fiber layouts. Utilize random forest/ensemble learning to consolidate multiple learners into final model-generated fiber layout result.

Achieved analytically optimal result using randomly generated set of less than <0.001% of all possible layouts. 

In collaboration with Anthony Zhou. Linked here are the final presentation and paper.

Designed 2D omni-directional drivetrain in Solidworks with five 4" omni wheels on aluminum and 3D printed frame elements, with bio-material based center-drop omni wheel to maintain constant ground contact. Created custom pen changer with mounting for three different pen color holders.

Integrated electronics developed on Arduino platform. Drivetrain powered by two NEMA 23 stepper motor connected to DM542T motor drivers, and color changer powered with servo.

Developed image processing software in Python for bluetooth-based wireless control scheme to communicate between computer and Arduino bluetooth module. Pseudo G-code is then processed onboard with Arduino code and sent to various motors for actuation.

In collaboration with Andrew Wang, Erik Francis, and Lekha Duvvoori. Link to final presentation here.

Designed 3-axis robot arm with carbon fiber rods and servo motors for datum positioning in composite molding. Printed mounting and interface components with Prusa i3MK3+ with PLA filament after design in Solidworks.

Created scaled 3-axis potentiometer setup to generate a PID control scheme, and implemented python control script with MQTT integration, allowing for long range control with minimal delay. Additional deadband controller for final angular control to mitigate noise in system.

 In collaboration with Erik Francis on project in both hardware design/manufacturing and software development/testing.