Using Game Theory to Improve Driving
“Game theory” has become a buzz word in a number of circles as people realize that they can utilize the same structure and reward systems present in games when building almost any system. This same line of thinking is starting to reach the automotive world where it could have a big impact on how we drive.
Game Theory – Displaying Good Driving Habits
Instant feedback tools showing a car’s current fuel consumption have long been a favorite of hypermilers, but one only needs to look at the short-lived fad for digital displays to understand how uncompelling a number on the dash is to the general public.
To that end, automakers have been working on displays that emulate the graphic rewards used in video games. Ford’s Fusion Hybrid used tree leaves and now butterflies to gradually change the look of the gauge cluster, Honda’s Insight shifts the speedometer from blue to green lighting, and the Prius c has a bar graph showing total fuel savings in dollars. Not only does it help overall fuel economy, it helps drivers learn the techniques needed to get the most out of hybrid and electric vehicles.
Welcome Your Robot Overlords: They Reduce Gridlock
Toronto is notorious for its traffic, but a new take on traffic control has alleviated some of that congestion using a novel approach to traffic light control.
University of Toronto Engineering graduate Samah El-Tantawy Baher and professor Baher Abdulhai have developed a traffic system called the Multi-agent Reinforcement Learning for Integrated Network of Adaptive Traffic Signal Controllers, aka “MARLIN-ATSC.” This new system acts like a multiplayer game of traffic management.
Instead of operating lights using a central system, each of the 60 intersections in the test had its own independent controller. El-Tantawy likens it to a soccer team: Each “player” is trying to score goals (move cars) by itself while also working with teammates to move cars through the entire area. The result is a system that can take into account the entire area, not just north-south and east-west movement. This decentralized approach can also adjust for malfunctioning players instead of bringing down the whole system in the event of an error.
The result? 40-percent less waiting time and 26-percent less total travel time. At just $20,000 per light, this system is gathering attention from road departments across North America as an inexpensive way to reduce congestion.