Roads

Note: this page (along with most of this site ;-) is still under construction.

My research on roads (in collaboration with some of my colleagues) is two-fold: one aspect is on automatic driving on roads, the other on analysing road related data to extract information about roads.

Automatic driving on roads

This is work done in collaboration with Elio Tuci, Seb Clarke and Marek Ososinski.

We have so far followed two different approaches. One is a simple, engineered and efficient method that models the road as a trapezoidal shape, using a single Gaussian to represent the colour statistics of the road area and the Mahalanobis distance to detect and track the road as the robot drives. This has led to two publications: a TAROS paper in 2012 and a paper to appear in the Journal of Field Robotics. Some of the data, including ground truth, used in the JFR paper is available from here.

The other approach uses evolutionary techniques to learn a neural system that analyses images and produces a control vector used to drive a (simulated for now) robot. A paper on this work is in the Proceedings of the Twelfth European Conference on the Synthesis and Simulation of Living Systems. We recently obtained funding from HPC Wales/Fujitsu for a PhD studentship to continue this work. No doubt we'll have more to add later.

Road data analysis

This is work we recently started in collaboration with Costain. We analyse image and laser data to extract assets pertaining to roads (road markings for now). Due to commercial agreements, we can't write more than that for now.