Ambarella, Inc. is a semiconductor design company that focuses on the development of low-power, high-definition, and ultra-high definition chips for image processing. Their chips are used in many types of imaging devices that require accuracy, precision, and detail and that are time-critical. Their chips are in wearable cameras, sports cameras, and automotive cameras. Ambarella is known for system-on-chip design, which refers to the high resolution and 4K processing speeds. In January 2017, Ambarella released its automotive multi-camera chips that provide the powerful processing performance that self-driving cars need in their cameras for capturing information about the immediate environment.
Ambarella’s Automotive Chip Technology
Ambarella has developed an automotive camera system-on-chips for use in self-driving cars. The chip features 4K resolution and ultra-high definition display. Ambarella touts its technology as a single-chip solution for autonomous cars that have single or multi-camera electronic mirrors. The chips allow for the capturing of a 360-degree surround view and video recording of the environment around the vehicle. Also built into the chip is an 800MHz dual-core central processing unit that uses Ambarella’s proprietary design technology. This central processing unit allows it to be used in autonomous cars that have driver-assist technology, wireless networking, and a user interface. The chip can handle up to three side-view cameras and one rear-view camera simultaneously. Because traditional cameras are challenged by dim and dark lighting conditions, Ambarella has also included an LED setup into the chip.
What Ambarella’s Technology Does in Autonomous Cars
This chip enables a self-driving car to perform multiple functions as it gathers the data from the immediate environment. For example, the chip could supply the necessary information for the vehicle’s autopilot to park the car without a human driver having to do it. According to Fermi Wang, who is the president and CEO of Ambarella, this chip technology presents a “unique architecture and a high level of system integration simplify automotive camera designs and lower overall system cost.” Lowering the cost of the technology that is built into autonomous cars is key to making the vehicles commercially viable.
Partnerships With Automakers and Other Self-Driving Car Technology Developers
Ambarella is working on its technology in conjunction with Gentex. The Gentex company is a supplier of rear-view mirrors that are able to integrate the display from a rear-view camera into the glass of the mirror. This partnership could allow for further enhancement of the digital display features for the passengers of the autonomous car.
How Ambarella’s Technology Works in Self-Driving Cars
In March 2018, Ambarella tested a self-driving car that makes use of its vision-processing chips. The test car uses stereo-vision cameras and computer vision algorithms for the processing of information in real time. The cameras and chips are designed to handle information in low-lighting and a high-dynamic range. This means that the autopilot could detect shadows in high-contrast scenes. The 8-megapixel resolution allows the artificial intelligence system to detect objects and potential hazards at a greater distance than the cameras and sensors that have been installed on other types of self-driving cars. Those algorithms allow for higher accuracy in processing information and reduce the number of chips required to handle the information from all the cameras.
Testing Ambarella’s Chips on Real Roads
The testing of Ambarella’s chips has taken place on a Lincoln MKZ that is outfitted with six cameras. The cameras allow for long-range detection of objects, road signs, weather, and other vehicles. The cameras are arranged in a hexagonal configuration on the top of the autonomous car, and each of the cameras provides a 75-degree view of the environment. The overlap allows the artificial intelligence system to sync the images and analyze them with accuracy and precision. Ambarella claims that this setup of cameras and the processing chip can detect a pedestrian from as far away as 150 meters. If the autonomous car were set up with additional neural networking software, that detection distance could increase to 180 meters. To complete the sensor, camera, and detection setup, the test vehicle also has a forward-facing radar and a computer for planning out the vehicle’s path. The radar engages in times of poor visibility and for redundancy of the images captured by the array of cameras.
Future Developments for Ambarella’s Technology
Ambarella has recently launched a second version of its chip. The newer version offers 20 times more processing power while only drawing 4 watts of power. The company aims to have the chip running all the self-driving car’s stack of technology on its own without the need for an additional PC. This sort of development could decrease the cost of autonomous cars, making them more feasible for commercialization. The tests of the autonomous vehicle showed a few issues with path-planning, so further GPS integration refinements are in the works.