Revision history [back]

For what you're asking it to do that looks about right. Unfortunately tf doesn't have full support for general relativity solutions.

But I would suggest that you rethink about what you're asking for. If you do truely need the global position with 6cm with less that 100ms latency this is what you need to do. But if you step back and think about your problem it's unlikely that you need super precise global position with very low latency. In general global fixes are moderately high latency and often are discontinuous.

As such the standard techinque is to decouple the low latency high speed positioning requirements and do all of those computations in the odometric frame. And then only deal with global positioning at a slower rate updating setpoint and goals in the odometric frame when new global information is received it updates the map -> odom transform with any drift observed. But between updates the only errors is the accumulation of drift in your odometry, which should be moderately small if you have a global fix streaming in, even if it's latent by 60ms and coming in at 10Hz.

For reference also: https://www.ros.org/reps/rep-0105.html