Keeping a Fresnel concentrator pointed exactly at the Sun is a filtering problem: raw sensor readings are noisy, dust adds bias, and every unnecessary actuator move costs power. This dashboard compares three Kalman-family estimators — linear KF, Extended KF, and Unscented KF — for lunar sun tracking so you can pick the one that maximises net power.
It visualises estimated versus true elevation and azimuth, tracking efficiency over time, and the compute and movement overhead each filter incurs.
A Kalman filter blends a motion model with noisy measurements to estimate the true sun angle. The linear KF is the cheapest baseline; the EKF linearises the nonlinear azimuth/elevation dynamics with a Jacobian; the UKF propagates sigma points through the nonlinearity for better accuracy under heavy noise.
The dashboard scores each on tracking efficiency minus the power spent computing and moving, because a filter that points slightly better but never over-corrects usually wins the net-power contest on a tight lunar budget — often the UKF.
Under realistic dust and noise, the UKF typically holds the focus a fraction of a degree tighter than the EKF while triggering fewer corrective moves, so its net efficiency edges ahead even after its slightly higher compute overhead is subtracted.
KF is linear; EKF linearises with a Jacobian; UKF uses sigma points for nonlinear dynamics and is usually the most robust here.
On a power-limited rover, an estimator that constantly actuates can lose more energy than it saves; net power is what matters.
It mirrors the same filter math a flight controller would use, at an educational fidelity.
UKF is the recommended default for this dust/noise regime; KF or EKF suit ultra-tight power envelopes.
Yes — browser-only, 25 languages.