This video shows world view of 5 sensors moving through an artificial environment in which observations are generated by a Gaussian process with predefined hyperparameters. Superimposed is a contourplot of the predictive variance with which the sensors can predict measurements throughout the environment. The lower the predictive variance, the better measurements at uninstrumented locations can be predicted. These predictions are made using the sensor's GP, based on measurements collected on the sensor's path.
All Comments