The Waymax simulator could help unlock improvements in reinforcement learning, leading to AV systems displaying emergent behavior. Waymo also plans to use the new simulator for future challenges to gauge the AV industry's advancement on certain problems and compare its technology. The company believes that these challenges can attract talent to the field of AV and robotics research.
Key takeaways:
- Waymo has launched a new simulator, Waymax, to train intelligent agents for autonomous vehicle systems. The simulator provides an environment for training, complete with prebuilt sim agents and Waymo perception data.
- The Waymax simulator is lightweight, focusing more on complex behaviors among multiple road users than on realistic-looking agents and roads. It is now available on GitHub for non-commercial purposes.
- Waymo regularly hosts challenges for researchers to help solve problems relevant to autonomous vehicles. The company plans to rerun a challenge next year using the new simulator to see how advanced the AV industry is on certain problems and how Waymo’s tech compares.
- The Waymax simulator could help unlock improvements in reinforcement learning, leading to autonomous vehicle systems displaying emergent behavior. This could result in safer autonomous driving.