A modified stochastic gradient descent algorithm for view-based SLAM using omnidirectional images.

•Modification of SGD (Stochastic Gradient Descent) algorithm for a SLAM problem.•Compact map representation, consisting of a reduced set of omnidirectional views.•A single-computation for the robot’s localization.•Adaption to work with omnidirectional constraints into the SGD.•Improved efficiency results by considering several constraints simultaneously.