Button Button Button

Section 3: Mobile Robot Positioning

Many vehicle tracking systems use GPS as the primary sensor and a gyro plus odometry as a secondary sensor. The latter helps compute the vehicle’s position during GPS outages.
The problem with this approach is that low-cost gyros have high drift rates, which result in large heading errors after just a few seconds or minutes.  
Heuristic Drift Reduction (HDR) estimates the momentary bias drift of the gyro based on tested and proven heuristics, and then subtracts that estimated drift from the gyro's output.
The result is a dramatic reduction of the effective drift, which results in up to two orders of magnitude lower heading errors!

HDR requires no hardware at all. It is a small software segment that can be added to any existing program code.

HDR  works with any gyro or IMU; it is most effective with low-cost/high-drift gyros.

Plot on the right: Experimental results from a highway drive with a Memsense nano-IMU in a small SUV. Total distance: 50 miles, total travel time: 48 minutes.  The drift rate of the gyros in the nIMU is >80 deg/hour. Within a few minutes the gyro data becomes useless (see red curve). The average heading error of the uncorrected gyro was 153 deg. In contrast, with HDR correction (see green curve) the average error was only 5 deg. This is a 33-fold improvement!

1. Ground truth was provided by GPS (cyan curve).
2. HDR parameters were tuned for best performance in this experiment.
3. With untuned parameters, HDR improvements range from 3-fold to 100-fold, depending on the drift rate of the gyro.
4. Odometry data, needed for plotting position, was derived from GPS data since we had no access to the odometer in the test car. 

Click on the "Papers" button at the top to see more experimental results.

Click on the "Video Clip" button at the top to see a fly-over animation created from Gyro+HDR and odometry data only.