The tradeoffs, dead-ends, and what held.
Posts cover control, vision, actuation, and mechanical design — written for engineers who want the reasoning, not a summary of the result.
Browse by domain or search for a specific problem — each category maps to a distinct area of the build stack.
PID tuning on a $15k arm: iteration log
Simulation gave a stable gain set. Hardware disagreed immediately. This post walks through six tuning sessions, what each oscillation told me, and why I changed the derivative filter mid-project.
Depth estimation without a depth camera
One capacitor, three weeks of noise
A motor controller timing fault that lived in hardware, not firmware. Tracking it down required scoping every stage — and rethinking how I read datasheets under load.
Monocular depth on a constrained compute budget. I compare three approaches, show where each breaks, and document the one that held up in field conditions.
Bracket geometry and why I iterated four times
State estimation when sensors disagree
Kalman filter implementation on an IMU and encoder pair with conflicting outputs. Covers the math, the failure case, and the sensor trust weighting that resolved it.
Load path analysis caught a stress concentration the simulation missed. Each revision is documented with the specific failure mode that forced it.