How does sensor fusion affect the SLAM process? If we have a stereo camera with GPS and IMU, how does sensor fusion benefit the SLAM? Do we use sensor fusion to enhance the localization phase of SLAM, and then the map generation becomes more accurate?
Sensor fusion is a process where you are gathering information from one or more sensors and inputting that data to your processor to process the data to make sense whether it is useful or not.
In your case, you have a stereo camera, IMU and GPS. If your question is whether all of these sensors altogether will increase the accuracy of the map, then the answer is no. The stereo camera alone can produce good quality 3D point-clouds/maps, adding the IMU and GPS is not going to increase the accuracy of the quality of the data gathered from the stereo cam. But the GPS and IMU along with the stereo cam will help you increase the accuracy of the localisation.
It also depends on what sensors you are using and their quality.
TIP: If you are using a stereo camera on your robot/drone and planning to put in dark environments, or places with low light, add a few lights to light up the area which is being scanned by the stereo camera.