The VSI crew has performed research in autonomous mobile systems since about 2002, starting with the construction of custom build RON.
From the very beginning, our robots have been equipped with cameras used for the perception of their environment.
A very substantial development in 2007 dealt with omnidirectional vision using an ultra-wide angle lens mounted face-up on a robot platform. In a series of papers, we described how the images obtained from a upwards – looking semispherical field of view can be used to steer a robot along a path of waypoints, or to steer it into a certain position in a known space.
Currently, we are investigating the usage of modern machine learning methods, such as deep learning and reinforcement learning to steer robots in multi-agent situations and cars in real traffic. This research is based on both on physical robots, as well as an simulations in large scale artificial environments.