Capable AI and Robots Will be Able to Leverage Super precise cameras and sensors to lessen software challenges

Super multi-gigapixel cameras and micron precision 3D motion detection and other advanced sensors will provide the awareness of the surroundings to make it easier to program robotics and artificial intelligence to perform useful tasks. Humans can get by with less precise vision and sensors to competently perform tasks like walking around buildings and driving cars. Super-sensors will make it easier for developers to make robots and AI more competent.

The self-driving car developed by Google uses expensive sensors such as LIDAR to help the robotic car to get real time understanding of the environment in which it is driving. Alternative super-accurate sensors could be more affordable alternatives. For different situations and applications the different cameras and devices and sensors could be the enabling technology for feasibility or to achieve cost targets for commercialization.

Different devices could make the programming problems more solvable. Currently the bottlenecks are around software complexity and affordability.

Yesterday we covered the scientific paper that described how the parallel camera approach uses off the shelf technology and will lead in the near term to 50 gigapixel point and shoot cameras.

Leap Motion has a 3D motion detector with ten micron precision.

Breakthrough that will lead in a few years to affordable 50 gigapixel cameras

Technology Review – The AWARE camera has 98 micro-cameras similar to those found in smart phones, each with 10-megapixel resolution. By positioning these high quality micro-cameras behind the lens, it becomes possible to process different portions of the image separately and to correct for known distortions. “We realized we could turn this into a parallel-processing problem,” Brady says.

The corrections are made possible by eight graphical processing units working in parallel. Breaking the problem up this way allows more complex techniques to be used to correct for optical aberrations.

The prototype camera capable of capturing and processing an entire image in just 18 seconds.

Eventually, as computer processing power improves, the hardware needed for such a camera should shrink. Portable gigapixel resolution could be useful in a number of ways. For example, additional pixels already help with image stabilization. “Also, if you increase the resolution, you increase the chances of automated recognition and artificial intelligence systems being able to accurately recognize things in the world,” Nourbakhsh says.

As graphical processors improve, so too will the speed of the camera, says Bradley. And although the prototype currently stands 75 centimeters tall–about the size of a television studio camera—the device’s size is dictated in large part by the equipment needed to cool the circuit boards.

“In the near term, we think this concept of a micro-camera imaging system is the future of cameras,” says Brady. By the end of next year, his group hopes to be able to produce and sell 100 units a year, each costing around $100,000. This is comparable to the cost of a broadcast TV camera, he says.

Gigapixel cameras could eventually allow events to be covered in new ways. “Rather than showing a camera angle that the producer lets you see, the viewer will be able to see anything in the scene that they want,” Brady says.

If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks