State Observability through Prior Knowledge: Tracking Track Cyclers with Inertial Sensors

State Observability through Prior Knowledge: Analysis of the Height Map Prior for Track Cycling

“…Since it does not increase, the estimate error is drift-free. It performs better than the previous estimator [11], which had an RMS of 0.096 s.…”

“…The average standard deviation decreased from 0.077 to 0.048 s compared to the previous estimator [11]. Hence, the precision of the estimator has greatly increased by changing the body frame.…”

“…Motivated by the success of prior knowledge in pedestrian tracking [2,6,7] and vehicle tracking [8][9][10] we conducted an experiment to track cyclists (see Figure 1) with an IMU as the only sensor [11]. The estimated trajectory was drift-free in three 20 min trials with an RMS of about 1.15 m. This paper extends the experimental work [11] with an observability analysis and an improved estimator with an RMS of about 1.08 m.…”

“…Ref. [11] The optimization problem as a factor graph. ac k and gy k are the inputs, x k the state, g k the dynamic model (8), a k the acceleration function (7), ω k the angular rate function (6), f v k the forward velocity prior (9) and h k the height-map prior (10).…”

“…Different surface types are analyzed to reflect the two major parts of the race track, the curves and the straight parts. In addition, an improved version of the original trajectory estimator [11] is presented, which takes account of the bike dynamics [15]. The estimator is evaluated on the original dataset.…”