Algorithms

Sensor fusion sketch

Case study preview

IMU + wheel odometry blending; where noise models helped and where they did not.

Problem

Wheel odometry drifts; the IMU is noisy on this platform. The question was how much blending buys you before complexity outweighs gains.

We tried a straightforward complementary-style blend with conservative gains, then compared against raw odometry on a taped course.

Noise models helped set expectations; the bigger win was detecting when GPS-denied segments needed a different trust profile.