Small robots make trade-offs obvious. Power budgets, communication range, and sensor quality all fight for the same cubic centimeters. Here are a few notes I keep returning to:
- Power is the real currency. Every milliamp-hour counts, so I model power draw per subsystem and keep a running budget. Having this on paper prevents late surprises.
- Sensing beats speculation. Before adding clever algorithms, I invest in good calibration steps and sanity-check scripts. Clean signals are worth more than a new model architecture.
- Tight loops win. When control loops stretch beyond tens of milliseconds, behavior drifts. I try to keep the critical path lean and push everything else off that path.
- Logs are a design tool. A tiny, timestamped log stream (even at 1 Hz) reveals failure modes faster than intuition. It also teaches me where to place future instrumentation.
These notes are snapshots; they will evolve as projects grow. My aim is to keep the robots honest to physics while layering in intelligence carefully.