Liam JenningsI’m Liam — based in Seattle, currently pursuing my Master’s in Robotics Engineering at WPI. I work on embedded systems, real-time controls, and hardware validation — and I like making things well.
Hybrid A*, PRM pursuit, and trajectory planning across three graduate projects, each with a full technical report.
coursework
Three projects from RBE 550 at WPI: Hybrid A* search for four vehicle models — from a holonomic point robot up to a car with trailer — with Reeds-Shepp connections; and a pursuit simulation pitting a grid-based A* agent against a firetruck planning over a PRM. Reports written to actually explain the algorithms, not just log results.
Hardware test engineering for autonomous warehouse robotics: HIL fixtures, validation, and automated tooling.
industry
Owned end-to-end hardware validation for embedded and electromechanical subsystems. Redesigned a hardware-in-the-loop traction motor fixture with closed-loop PID and MQTT-over-Ethernet for multi-month accelerated life testing, and built Python tooling for log parsing and time-series visualization with InfluxDB.
Embedded electronics for WPI’s competition EV: the inverter controller and a driver radio telemetry board.
open source
Designed and validated boards for the 2024–25 car: an STM32 mixed-signal inverter controller that conditions pedal inputs and commands the motor controller over CAN, and a Bluetooth driver-radio board for live telemetry — PCB design through firmware and vehicle integration.
Sliding mode control for a quadrotor autonomously chasing down an intruding UAV in simulation.
coursework
A sliding mode controller regulating altitude, roll, and pitch, paired with trajectory prediction of the target, so the quadrotor can track, intercept, and return with an intruder inside a 10 m bounded airspace — staying stable through the disturbance of capture.
A planar arm designed, analyzed, and actually fabricated to move a 500 g payload between poses.
coursework
Full-stack arm design for RBE 501: forward kinematics via the Product of Exponentials, closed-form geometric inverse kinematics, trajectory generation, dynamics analysis, and motor selection. The course only required the analysis — we built the physical arm anyway.
A convolutional variational autoencoder that removes raindrop blur from photographs.
open source
A CVAE trained to reconstruct clean images from rain-degraded inputs, built as a machine learning course final project. The write-up walks through the architecture, the DeRaindrop dataset, and an honest look at where the reconstructions fell short.
Every write-up on the site, including the showcased projects above.
From a holonomic point robot to a car with a trailer — how the same search adapts as the kinematics get harder.
Closed-loop PID, MQTT over Ethernet, and what it takes to keep an accelerated life test honest.
An STM32 inverter controller and a driver radio board for WPI’s 2024–25 car — PCB layout through firmware and integration.
A quadrotor tracking, capturing, and returning with an intruder in simulation — and staying stable through the capture.
Kinematics, trajectories, dynamics, and motor selection for RBE 501 — plus the fabrication the course didn’t require.
A variational autoencoder trained to remove raindrop blur, and an honest account of where the reconstructions fell short.
Teaching a simulated TurtleBot to push a cylinder to a goal in Gazebo — and what the reward function got wrong first.
Rangefinders, encoder kinematics, and an OpenMV camera, combined to navigate a random maze autonomously.