Racing cars using machine learning

August 2020 - Our collaboration with Amazon Web Services (AWS) is changing how we use big data in the workplace. And what better way to test the technology than create a company-wide game?

Twenty-one staff from across Cenovus met virtually to race cars using AWS DeepRacer. The machine learning service uses reinforcement learning to achieve a specific goal – in this case to complete three laps around a track at top speed while avoiding crashes. Before the race began, participants had to train the program to steer, accelerate, brake and turn their car using a reward system similar to what is used to train animals. After receiving initial commands from the ‘driver,’ DeepRacer works independently to control the vehicle.

“DeepRacer helped me learn about AWS tools in a fun way while applying Artificial Intelligence to solve new types of problems,” said Thiago Avila, Staff Optimization Engineer and race winner. “It was exciting to see the variety of innovative approaches used to try to get their car across the finish line.”

Cenovus senior leadership also tried their hand at the race. Kudos to Harbir Chhina, Cenovus Senior Vice-President & Chief Technology Officer, who made it into the top five!

After the race, competitors discussed other ways smart machines could be used across Cenovus, and the oil and natural gas industry, including potentially predicting equipment failures or process upsets, modeling reservoir optimization or estimating future emissions.

"We’re trying new ways to improve our business performance through the power of information and data,” said Sarah Walters, Cenovus Senior Vice-President, Corporate Services. “This was a fun activity to get staff involved and see the capability of reinforcement and machine learning."