Texas Aerial Robotics, UT Austin
Thrust-Vector-Controlled Dart
Feb 2026 — Present
PID + iLQR control stack that steers a powered dart to a precise landing point.
The problem
A thrust-vector-controlled dart is an inherently unstable system: a rigid body balancing on its own thrust, with a fast-rotating gimbal as the only control input. Getting it to not just stabilize but fly to a chosen landing coordinate requires a controller that understands the full nonlinear dynamics — a simple PID loop alone won’t cut it.
The control stack
I built a three-part control system combining PID loops for fast inner-loop attitude stabilization with an iterative Linear Quadratic Regulator (iLQR) for trajectory optimization. iLQR repeatedly linearizes the dart’s dynamics around a candidate trajectory and solves the resulting LQR problem, converging on a locally optimal path to the target landing point. The PID layers then track that trajectory against disturbances the optimizer didn’t model.
Testing without crashing
Flight testing an unstable vehicle is expensive in airframes. I’m constructing an orbital test rig that constrains the dart to a safe envelope while still exercising the full control loop — letting us measure controller consistency across repeated runs and collect the data needed to refine the dynamics model.
Why it matters to me
This project is where my math actually fights back. Optimal control is beautiful on paper; making it survive sensor noise, actuator lag, and imperfect dynamics models is the engineering part — and that gap is exactly where I want to work.