For the purpose of developing an autonomous terrestrial vehicle the ESL uses a Kymco MXU 150 ATV. This quad bike has been modified to allow full drive-by-wire control by an on-board computer. The quad bike is fitted with an array of sensors including GPS, accelerometers, gyroscopes, magnetometer, a light detection and ranging (LIDAR) sensor and two monochrome cameras.
Mapping is one of the focus areas relating to the quad bike. Determining the position and orientation of the vehicle is critical for an autonomous vehicle. The vehicle states are estimated using a kinematic estimator, which uses accelerometer, gyroscope, magnetometer and GPS measurements. Work is underway to improve these state estimates using stereo vision based SLAM (simultaneous localisation and mapping).
In order for the vehicle to safely move around in its surroundings, it needs an accurate dense map of its environment. We are looking at various sensors and mapping techniques to this end. Currently, the LIDAR sensor measurements are used to build a 3D occupancy grid map of the vehicle environment. We are working on combining the LIDAR and stereo camera measurements to build the occupancy grid map, as well as improving the efficiency of the mapping algorithm by using adaptive occupancy grid mapping. The last focus area of our research is control and guidance. Once the vehicle has constructed a map of its surroundings, it can start planning routes through its environment. To ensure that the planned routes stay collision free, we also concentrate on detecting possible collisions and how to resolve them. Current work focuses on path planning for partially mapped environments.