Revision history [back]

There is no way to eliminate drift, but there are some things you can always do. Dont know what kind of gyro you are using, but for raw sensor you can: - compensate with calibration, collect some amount of samples while gyro is as stady as possible and avarage of measurment use as calibration value (it will change with temperature and/or voltage, depending on the sensor) - use magnetometer for Z axis and/or accelerometers for XY, but it makes it INS and gats more complicated (and way more expensive) - if robot is not moving, you can detect no signals from wheels at all. That makes odometry way more reliable than in motion, so you can disregard gyro readings. It may improve Z accurqacy if robot is in stop for a long time. It will mess things up if you roatate robot without moving the wheels though. - SLAM algoritms may help, if you plan to use Kinect, thats the way. Read about rgbslam and Kinect, materials on it are all over the web (including ros wiki and ros answers)

There is no way to eliminate drift, but there are some things you can always do. Dont know what kind of gyro you are using, but for raw sensor you can: - can:

• compensate with calibration, collect some amount of samples while gyro is as stady as possible and avarage of measurment use as calibration value (it will change with temperature and/or voltage, depending on the sensor) - sensor)
• use magnetometer for Z axis and/or accelerometers for XY, but it makes it INS and gats more complicated (and way more expensive) - expensive)
• if robot is not moving, you can detect no signals from wheels at all. That makes odometry way more reliable than in motion, so you can disregard gyro readings. It may improve Z accurqacy if robot is in stop for a long time. It will mess things up if you roatate robot without moving the wheels though. - though.
• SLAM algoritms may help, if you plan to use Kinect, thats the way. Read about rgbslam and Kinect, materials on it are all over the web (including ros wiki and ros answers)

There is no way to eliminate drift, but there are some things you can always do. Dont know what kind of gyro you are using, but for raw sensor you can:

• compensate with calibration, collect some amount of samples while gyro is as stady as possible and avarage of measurment use as calibration value (it will change with temperature and/or voltage, depending on the sensor)
• use magnetometer for Z axis and/or accelerometers for XY, but it makes it INS and gats gets more complicated (and way more expensive)
• if robot is not moving, you can detect no signals from wheels at all. That makes odometry way more reliable than in motion, so you can disregard gyro readings. It may improve Z accurqacy if robot is in stop for a long time. It will mess things up if you roatate robot without moving the wheels though.
• SLAM algoritms may help, if you plan to use Kinect, thats the way. Read about rgbslam and Kinect, materials on it are all over the web (including ros wiki and ros answers)

There is no way to eliminate drift, but there are some things you can always do. Dont know what kind of gyro you are using, but for raw sensor you can:

• compensate with calibration, collect some amount of samples while gyro is as stady as possible and avarage of measurment use as calibration value (it will change with temperature and/or voltage, depending on the sensor)
• use magnetometer for Z axis and/or accelerometers for XY, but it makes it INS and gets more complicated (and more expensive)
• if robot is not moving, you can detect no signals from wheels at all. That makes odometry way more reliable than in motion, so you can disregard gyro readings. It may improve Z accurqacy if robot is in stop for a long time. It will mess things up if you roatate robot without moving the wheels though.
• SLAM algoritms may help, if you plan to use Kinect, thats the way. Read about rgbslam rgbdslam and Kinect, materials on it are all over the web (including ros wiki and ros answers)