Hungarian company AImotive is working to develop a more affordable autonomous driving system, at their California office. Most other self-driving projects utilize ‘Lidar,’ an expensive system, similar to radar, that allows for depth perception and obstacle detection. AImotive is working to accomplish the same goals using normal cameras working alongside artificial intelligence. The project would allow the company to convert cars for around 6,000 dollars, compared to the 70,000 to 100,000 necessary for standard methods.

CEO and founder Laszlo Kishonti said: “The whole traffic system is based on the visual system. Drivers don’t have bat ears and sonars, you just look around and drive.” To simulate the human ability to drive, “The only way to do this is with AI,” he added.

The company has used four fish-eye cameras on each side of their vehicle, in addition to dual stereo cameras on the front and back of the vehicle. Their car uses a powerful PC, placed in the trunk, to stitch together the feeds from the cameras in real time, creating a three-dimensional model of the car’s surroundings.

AI is used to make decisions about the surroundings, based on the nature of the environment and the surrounding objects. Such tasks pose a challenge for AI, despite being simple for humans.

The car combines its live view from the cameras with a GPS “location engine” that situates the vehicle on a map. An AI-driven “motion engine” controls the cars movement on the road, and feeds into a “control engine” that determines steering, acceleration, and braking.

While the company does not yet have its California driverless car license, their modified Prius has been tested on roads and car parks in Budapest, where AImotive has its headquarters.

Lidar has been the industry standard system for driverless vehicles since it performed well in DARPA’s driverless car competition 10 years ago, according to Kishanti.

AImotive was able to take advantage of a surge in research when the company launched in 2015, according to COO Niko Eiden.

The vision based approach offers both limitations and advantages. Their system, like humans, does not perform as well when vision is limited, like in fog or snow. However, the approach could prove to be more flexible, able to adapt to new areas and respond well to unpredictable changes in the route, such as redirected traffic.

Currently, AImotive is not producing its own cars from scratch, but is working with companies like Volvo to provide driverless systems for cars and trucks. The company is also focused on getting its California driverless license, and will begin testing its system in urban settings in early 2018.

Leave a Reply

Your email address will not be published.