Intel and collaborators won a defense grant to develop advanced simulations for off-road autonomous ground vehicles, the company announced Tuesday.
The Defense Advanced Research Projects Agency award is for research in the program called Robotic Autonomy in Complex Environments with Resilience – Simulation, or RACER-Sim. Intel is collaborating with Computer Vision Center in Barcelona and the University of Texas at Austin.
The simulation is expected to reduce the risks and costs of the off-road autonomous vehicle development as compared to traditional testing and verification. Intel released a video showing real-world vehicle crashes during testing. By contrast, a simulation crash would be far less costly and dangerous
Intel noted that many vehicle simulations exist, but few are available for off-road autonomy development. Real-world testing is now the primary method being used.
Off-road vehicles need to be tested on extreme terrain with rocks and all kinds of vegetation, and the RACER-Sim approach cuts into the time required. It will be conducted in two phases over 48 months. Intel Labs will use its simulation platform to customize future maps covering more than 100,000 square miles with just a few mouse clicks. Autonomous vehicle simulations typically test virtual sensors positioned around vehicles to record and refine what they would see or record with cameras, radars and lidars.
Also Intel and collaborators will also develop new algorithms without using a physical robot, validating its performance in simulation. They are developing new sim2real techniques to train the robot in simulation to acquire skills and transfer the skills to a corresponding real robotic system.
The goal is to create a “versatile and open platform to accelerate program in off-road ground robots for all types of environments and conditions,” said German Ros, director of the autonomous agents lab for Intel Labs.
RELATED: How GM, Honda plan to chase the golden ring of sub-$30k EVs