allow pause on saving clouds - #14
Conversation
|
|
||
| Measurement::Ptr MeasurementStorage::get(const boost::uuids::uuid& uuid) | ||
| { | ||
| return mMeasurements.at(uuid); |
There was a problem hiding this comment.
I don't agree with this change. When you use get(), you should catch and handle std::out_of_range, not be forced to check the returned pointer.
There was a problem hiding this comment.
We can change that, imo there is not much difference.
Currently the behavior is similar to of std::map::get().
| void MeasurementStorage::add(Measurement::Ptr measurement) | ||
| { | ||
| mMeasurements[measurement->getUniqueId()] = measurement; | ||
| if (enabled) { |
There was a problem hiding this comment.
I am not sure this pausing function should go into the MeasurementStorage. It feels kinda wrong to call add() and then have this method silently not add the measurement because of some internal state. Wouldn't it be better to do this on a higher level (node/application) and just not call MeasurementStorage::add() when in paused mode?
There was a problem hiding this comment.
Yes, I also thought about this, but it requires a lot of changes in the other slma3d classes, as there still should run a icp and a vertex should be created, adding it here was the least intrusive way of having this funcionality.
There was a problem hiding this comment.
We could add a pause function to the graph class, but that info was to be passed down to all the implementations that use add() on the storage.
| // throw BadMeasurementType(); | ||
| // is there was an emptyx cloud, use the guess | ||
| Covariance<6> covariance = Covariance<6>::Identity() * mCovarianceScale; | ||
| return Constraint::Ptr(new SE3Constraint(mName, guess, covariance.inverse())); |
There was a problem hiding this comment.
Adding the guess (which is just the current relative position of the two vertices) as a constraint is a bad idea. A PointcloudSensor can only create constraints between pointclouds. So similar as above, when an application wants to do something else, catch the BadMeasurement and do everything needed in the handler.
There was a problem hiding this comment.
This code is only used, when the previous cloud was not saved ( slam was paused), so there is no cloud to compare to. The hope is that eventually a loop closure will catch this ans optimize correctly, as all the vertices in between are available
2 participants
This allows to call (disable) in a storage implementation, so that e.g. point clouds are not actually saved, the contraint is still created, but the cloud is not saved. On enable() the initial guess is trusted and mapping can go on.