AV driving on off-road terrain is more challenging than it seems. The AV needs to consider narrow trails, tunnels & bridges in order to validate the dimensions & angels needed for passibility, as well as detection and reaction to unexpected ends of drivable paths.
In addition, vehicle physics comes into play – the autonomous vehicle needs to calculate vehicle dynamics versus paths tilt angles while not having road marks on a flat road surface for detecting free space, using detected surrounding objects (or none) only. Cognata’s simulation platform enables the creation of many use cases that the autonomous vehicle might encounter and the validation (and analysis) of such cases.
Off-road sensor simulation is facing challenging suspension and manoeuvring, trying to overcome the simulated sensor’s movements as well as overcoming blind spots created by the need to have a sheltered sensor placement.
Cognata’s simulation platform can create multiple use cases with limited visibility by other objects, dust trails, vegetation, and ground blindspots as well as darkness, lights & shadows for validating driving situations with limited visibility and ground blindspots, in addition to teaching the vehicle to compensate for sensor movements due to a bumpy road.
Focusing on off-road AV training requires a special focus on two additional aspects: off-road scenarios and off-road assets.
For the scenarios catalog, we have created a variety of different scenes designed to train and validate off-road driving from identifying a drivable path to overcoming obstacles and difficult road conditions.
We have also created a rich catalog with off-road special synthetic assets such as people, obstacles, and special off-road elements.