The main goal of the CARLA Autonomous Driving Challenge , says the organization, is to achieve driving proficiency in realistic traffic situations and help expedite the race towards AV-based mobility. In the challenge, participants will deploy state-of-the-art autonomous driving systems to tackle complex traffic scenarios in CARLA.
The autonomous driving systems - referred to as "agents" - will demonstrate their proficiency by driving along complex urban and highway scenarios. Traffic situations were selected from the NHTSA pre-crash typology, which include negotiations at traffic junctions, dealing with pedestrians, lane merges, and more.
The challenge will evaluate how proficiently and safely agents address each situation – and consists of four parallel tracks, focusing on different possible configurations of AVs. Each of the routes will be repeated multiple times under different weather and traffic conditions. Routes will use a variety of maps, including freeways, urban scenes, and residential districts.
The performance of an agent will depend on the number of routes successfully completed. A route is considered successfully completed if no critical infractions were triggered. If the agent triggers a critical infraction, the episode will be automatically terminated and the agent will only get a score proportional to the percentage of the route it completed until the critical infraction.
For each track, three winners will be awarded monetary prizes. Teams from both academia and industry are encouraged to participate. Winners will be notified at the end of the competition and results will be publicly announced at the CARLA Autonomous Driving Challenge Workshop , held at CVPR 2019 in Long Beach, CA.
The challenge is sponsored by industry leaders AWS, Waymo, Uber ATG, Audi EV, and AlphaDrive.
Addressing the challenges of autonomous driving
Blockchain challenge looks for smart mobility solutions
Automated vehicle coalition formed to educate public
Mapping startup for autonomous cars aims to automate cities
Amazon scale-model autonomous car aims to jump-start devs with AI