There was a great turnout for the first AI Driving Olympics competition, which took place at the NeurIPS conference in Montreal, Canada on Dec 8, 2018. In the finals, the submissions from the top five competitors were run from five different locations on the competition track.
Our top five competitors were awarded $3000 worth of AWS Credits (thank you AWS!) and a trip to one of nuTonomy’s offices for a ride in one of their self-driving cars (thanks APTIV!)
Team Panasonic R&D Center Singapore & NUS
(Wei Gao)
Check out the submission.
The approach: We used the random template for its flexibility and created a debug framework to test the algorithm. After that, we created one python package for our algorithm and used the random template to directly call it. The algorithm basically contains three parts: 1. Perception, 2. Prediction and 3. Control. Prediction plays the most important role when the robot is at the sharp turn where the camera can not observe useful information.
Check out the submission.
The approach: “I tried and imitate what a human does when he follows a lane. I believe the human tries to center itself at all times in the lane using the two lines as guides. I think the human implicitly projects the two lines into the horizon and where they intersect is where the human directs the vehicle towards.”
Check out the submission.
The approach: “The AI-DO application I made was using the ROS lane following baseline. After running it out of the box, I noticed a couple of problems and corrected them by changing several parameters in the code.”
(Mikita Sazanovich)
Check out the submission.
The approach: “We used our framework for parallel deep reinforcement learning. Our network consisted of five convolutional layers (1st layer with 32 9×9 filters, each following layer with 32 5×5 filters), followed by two fully connected layers (with 768 and 48 neurons) that took as an input four last frames downsampled to 120 by 160 pixels and filtered for white and yellow color. We trained it with Deep Deterministic Policy Gradient algorithm (Lillicrap et al. 2015). The training was done in three stages: first, on a full track, then on the most problematic regions, and then on a full track again.”
(Anton Mashikhin)
Check out the submission.
The approach: Our solution is based on reinforcement learning algorithm. We used a Twin delayed DDPG and ape-x like distributed scheme. One of the key insights was to add PID controller as an additional explorative policy. It has significantly improved learning speed and quality
We’re just about at the end of the road for the 2018 AI Driving Olympics.
There’s certainly been some action on the leaderboard these last few days and it’s going down to the wire. Don’t miss your chance to see you name up there and win the amazing prizes donated by nuTonomy and Amazon AWS!
Submissions will close at 11:59pm PST on Thursday Dec. 6.
Please join us at NeurIPS for the live competition 3:30-5:00pm EST in room 511!
Submissions to the AI Driving Olympics are officially open for the Lane Following task evaluated in simulation.
To make a submission, follow the instructions in the AIDO-book.
For example, a good place to start is the Quick Start Guide.
For discussion please check out the AIDO forums.
The competition will use the Duckietown platform, a scaled-down affordable and accessible vision-based self-driving car platform used for autonomy education and research. This open-source project originated at MIT in 2016 and is now used by many institutions worldwide.
The AI Driving Olympics is presented in collaboration with 6 academic institutions: ETH Zurich (Switzerland), Université de Montréal (Canada), NCTU (Taiwan), TTIC (USA), Tsinghua (China) and Georgia Tech (USA), as well as two industry co-organizers: nuTonomy and Amazon Web Services (AWS).
The competition will comprise 5 challenges of increasing complexity: 1) Road following on an empty road; 2) Road following with obstacles; 3) Point-to-point navigation in a city network; 4) Point to point navigation in a city network with other vehicles; and 5) Fleet planning for a full autonomous mobility on demand system.
Competitors will have access to simulators, logs, reference implementations, and finally real environments (“Robotariums”) that will be remotely accessible for evaluation. The entries that score best in the robotariums will be run during the live event at NIPS 2018 to determine the winners.
The competition aims at directing academic research towards the hard problems of embodied AI, such as modularity of learning processes, and learning in simulation while deploying in reality. The competition also promotes the democratization of AI/robotics research by offering a common infrastructure available to everybody through the use of remote testing facilities.
Competitors can also build their own Duckiebots using provided DIY instructions, or buy Duckiebots and Duckietown hardware through a kickstarter campaign.
For rules and timeline, please see the site https://driving-olympics.ai/