In this robot mower simulation we are trying to 'map' and 'localize' (position error < 0.5m) - but just with odometry, a perimeter boundary,
a compass (including gyro+accel) and GPS. Do you think it's possible?
The driven distance is computed by odometry sensor, the direction by compass. The short-time error is corrected
when the robot hits the perimeter boundary. The long-time error is corrected by GPS (if the computed position is
off by more than 4 meters).
Any ideas how this could be improved, would it work better using Kalman filters maybe?
- real world:
- sensors simulated: odometry, compass, GPS
1. odometry sensor error (due to slippage) 5-30% per meter
2. GPS position error < 3.5m
3. compass error < 0.5 degree
- virtual world (the robot builds itself)
- computed position using simulated sensors:
1. odometry: distance (short-time)
2. compass: direction (short-time)
3. GPS, perimeter: position recalibration (long-time)