DriveML Huawei Autonomous Vehicles Challenge

Overview

Have you recently been hearing about self-driving cars? Are you interested in applying your knowledge to an application that will make the world a better place? If so, Huawei's drive-ml competition is your opportunity to leave your mark on a thriving and booming research area holding the promise of a new generation in transportation development.

Intending to further research in autonomous driving and in identifying rising stars, we invite you to participate in a competition organised by a world-leading team of autonomous decision making.

Your goal is to design an agent that is capable of driving safely and efficiently across a variety of simulated maps. We offer you these scenarios through an in-house developed state-of-the-art simulator that emulates real-world behaviours at different granularity levels. Our simulator is the first of its kind in that it can model realistic dynamical behaviour getting us a step closer to bridging the gap between research and application.

Your challenge consists of two-phases covering both single and multi-agent reinforcement learning scenarios. Winners will be announced in a ceremony, with cash prizes of up to £5,000. You will also get the chance to teach us all about your innovations in a spot-light talk during the event. Even better, we offer you an opportunity to travel to Huawei headquarters in China to learn all about our cutting-edge research.

Gather your friends, form your team, and go ahead and register at drive-ml.com. Let’s get these cars moving!

For any questions do not hesitate to contact us at UKchallenge@huawei.com.

The Task

Competitors will train a driverless car from scratch to traverse a multi-lane highway as fast as possible whilst avoiding collisions with social vehicles. Success will require the training of intelligent driving behaviours which account for the behaviour of nearby vehicles, such as when executing an overtaking manoeuvre.

Observations, Actions and Rewards

Competitors will be provided with a continuous low-level action space of throttle, break and steer as well as the flexibility to build discrete higher-level actions on top of this as they see fit. They will also be provided with a range of base observations from which to choose from, including,

• Lane angle, distance from the center of the lane and ego vehicle state
• Relative lane positions and velocities of nearby vehicles
• Occupancy grid maps, which indicate the positions of nearby vehicles on a grid
• Top-down rgb images centred on the competitor’s vehicle

And also the freedom to design their own observations. They will also be able to shape rewards, such as by punishing collisions or incentivising lane speed, in order to facilitate training of effective policies.

Special thanks to the HiWay and DriveML development teams: David Rusu, Julian Villella, Peyman Yadmellat, Kasra Rezaee, Jun Luo, Yaodong Yang, Sanjeevan Ahilan, and Haitham Bou Ammar.