Avoiding dangerous situations that involve other road users is one of the primary challenges in the development of self-driving cars. The detection of nearby objects such as cyclists, pedestrians and other cars is a prerequisite for avoiding these situations. In addition to determining position, size, and orientation of these object, estimating their velocity is very important.
Aside from radars and cameras, LiDAR sensors can be used for obtaining information about object in the surrounding of the vehicle. LiDAR sensors scan the periphery of the vehicle using laser beams and create a three-dimensional representation offering higher resolution and accuracy than radar sensors.
As part of the project "Multi-Object tracking using LiDAR", methods for detecting and tracking multiple objects in the surrounding of the vehicle are being developed. These allow the combination of data from multiple LiDAR sensors and information from previous measurements to determine the object state including the object's size and velocity. By using probabilistic models, uncertainties in measurements can be considered explicitly and all availably information can be fused to create a representation of the vehicle environment with higher certainty.