You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The main requirements from our mapping system are -
The interface is probabilistic, using Bayes algorithm for successive updates. This'll be useful, as it'll take into consideration the wrong readings or fuzzy reading that may come up with the sensors.
The interface allows for multiple inputs simultaneously. This will be very useful since we'll be able to construct a much larger and more detailed map, using multiple sensors.
The interface is sufficiently optimised, to be feasible for usage on a mobile robot such as ours, where computability is important.
Now the first problem, can be solved, using most general mapping packages, which use probabilistic maps updated using Bayes algorithm. The second problem, also is easily solved because of the first one. Basically assigning two or more different inputs, both with their own trust values, will not effectively cause any problems and our map will be updated based on them both.
The third problem is the one we are working on currently. The current plan is to use a chunk loading system, like in games, where chunks of the map are successfully loaded and unloaded from the primary memory and stored away into the hard drive, as the robot goes away from them.
This is the main thing we are trying to integrate with the octomap way of storing of maps. We are making a new mapping system from scratch heavily based on the octomap system, but which does not store the entire map in RAM, and instead offloads chunks into the hard drive when necessary.
No description provided.
The text was updated successfully, but these errors were encountered: