ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
1 | initial version |
You can put them as zeros and give them large covariances. Yes they will be filtered, but to what end? I am not sure you have enough information (2D accelerations) to do much with. You will also need some method to sense orientation, which is the most important. That is usually done with gyros and a 3D accel or perhaps a camera (could give you a rough estimate).
If you are trying to do some strap on inertial nav stuff, this won't be good enough. The accel errors will grow without bounds and your pose will be junk. You said you don't have wheel encoders, but what about video odometry or GPS to attempt to correct the drift.