Autonomous vehicles don’t need drivers to control them. But to make these self-driving vehicles completely safe, they must be made equivalent to a human driver who has developed cognitive and decision making skills through experience. Now, a team of researchers from the Massachusetts Institute of Technology and Toyota has developed a system that ensures autonomous vehicles can safely navigate through traffic.
Traditional automated systems need direct visibility while navigating to avoid obstacles, so when the line of sight is broken the systems can fail. The team has developed it’s own model based on probability to calculate the potential risk of collisions or other traffic disruptions at intersections. The study on development was published in IEEE Robotics and Automation Letters.
The system considers factors like visual obstructions, sensor data, the speed of cars and the attentiveness of drivers to actively calculate the estimated risk of collision. Based on the results, the system will alert the car to act accordingly by stopping or pulling forward. The system has been designed to help autonomous cars navigate through busy intersections or maneuver a roundabout.
The researchers performed multiple trials of the system in a model city with busy streets using remote-controlled cars. The test consisted of autonomous cars and remote-controlled cars that were assisted by the system. The system was able to successfully help the cars avoid a collision, at least 70% percent of the time, depending on various factors.
According to the researchers, the system showed that it was efficient and fast enough to be used with an autonomous test car. The system will act as a supplemental risk metric that will provide better reasoning for autonomous vehicles. The system can also be used in semi-autonomous vehicles where drivers maintain shared control of the vehicle. The researchers will work on including other risk factors like the presence of pedestrians in their system.
Image Credit: MIT News