Machines master mist as MIT developers improve self-driving cars’ visibility
Though self-driving technology is becoming more and more sophisticated and extensive advances in the field are bringing a growing number of manufacturers around to the idea of dispensing with the driver and letting the car just get on with it instead, a few things have proved themselves too difficult for autonomous cars to tackle.
One of these has been adverse weather conditions, such as fog and mist, which throw off the all-important laser sensors these cars use for navigation, as the water droplets in the air cause the waves of light the vehicles emit to scatter, meaning cars can’t accurately distinguish obstacles or nearby surroundings (such as the edge of roads). In such circumstances, today’s self-driving cars nervously pass the task of driving back to the human driver, who takes charge of the car whilst needed.
However, developers at MIT in the USA have come up with new tech that will make cars’ perceptions equal to (or better than) humans drivers’ in low-visibility.
The developers created a code that compensates for the additional noise the fog produces, and tested it in a metre-long fog chamber, inside of which the system was capable of detecting the shape of wooden blocks, a human-shaped figurine and even letters that a human eye couldn’t see.
While this is a significant advance in the field, autonomous car pioneer and Tesla CEO Elon Musk has said that a sensor-based approach to driverless cars is limited and ‘a crutch’, advocating instead an approach based around image recognition and radar.
We’ll keep you updated on all the driverless news we come across!