A novel method of traversable area extraction fused with LiDAR odometry in off-road environment

In off-road environment, to ensure the safe driving of the unmanned-vehicle, we need to get the traversable area for the car. Compared to the urban area, there are not only positive obstacles, but also negative obstacles as well as cliffs in off-road environment. Our paper proposes a novel method to extract the traversable area for the unmanned vehicle. We fuse three LiDARs in order to detect different kinds of obstacles. What's more, we use the result of the LiDAR odometry, which has been proposed by our team, to transform the detection results based on the ego frame into the global one. After that, the historical traversable area is mixed together with the current results on basis of the Bayesian Theory. Ultimately, we get the current off-road traversable area. The method has been tested in off-road environment. The result proves that our system has strong robustness and it's less time consuming. The purposed method is able to get a reliable traversable area for unmanned-vehicles.