Kalman TVC
Kalman is a lightweight, self-stabilizing rocket designed to gimbal a high-impulse solid propellant motor, named for the mathematician whose filter makes that stabilization possible. It's a development platform, built to practice the full engineering cycle: requirements definition, analysis, fabrication, and test.
The project is organized around three big pillars: Analysis, Hardware, and Software.
(not sponsored by diet coke)
Some Background
The central engineering challenge in Kalman is the TVC mechanism, a two-axis system that has to transmit load, respond quickly, and introduce as little mechanical error as possible. The design has gone through several iterations, moving from a ball joint to a standard U-joint to a no-slip bearing U-joint, each step trading complexity for tighter constraint and reduced slop. Manufacturing tolerance is a real factor here.
When working with FDM-printed parts and hand-machined components, error will stack up, and the control system has to fight any slop that makes it through. The design process is just as much about understanding those interactions as it is about optimizing geometry.
On the software side, the control law is built as a linear quadratic regulator in MATLAB, developed from first principles using the vehicle's state-space model. The control law is then ported to ArduPilot via Lua scripting, which runs on the Pixhawk flight controller onboard the vehicle. Testing follows a hardware-in-the-loop approach: an MPU-6050 provides angular rate and attitude data from the physical gimbal, and a Claude Code agent runs collecting this data and converging over test cycles to define the response of the system. The goal before any static fire is a fully instrumented bench test that characterizes system response under real actuation.
The analysis work covers two primary concerns. The first is structural, as gimbal deflections under expected flight loads are modeled in ANSYS, margins of safety are defined for the current design, and the geometry is iterated to shed weight where the analysis shows margin to do so. Hand calculations from Shigley's run alongside the FEA as a sanity check, particularly at interfaces where reaction loads need to match intuition before trusting the solver output. The second concern is thermal. Solid propellant motors generate significant heat over their burn, and a long-burn motor will thermally soak the gimbal and into the lower thrust structure, which is a chopped carbon fiber reinforced polymer. The analysis defines the temperature distribution through that load path and confirms that no part of the structure reaches a temperature where material performance degrades enough to matter. Both analyses feed back into design decisions and, eventually, into how the software handles off-nominal behavior in flight.
Kalman pulls together work from several directions at once including fabrication and tolerance management from my machining and composites experience, structures and thermal analysis methods developed as an intern at SpaceX, and GNC theory applied in earlier projects like Laika and AETHER. The point of the vehicle is admittedly not the vehicle itself. It is the process of defining requirements, doing the analysis, building the hardware, closing the loop in software, and finding the problems I did not expect. This page is where I document that process!
