Mobileye Details "Human-like" Algorithmic Driving Platform for Autonomous Cars at CES 2017
【Summary】Mobileye, a company that specializes in optimized sensing platforms for driverless vehicles, has taken a completely different approach to such traditions by revealing various components of its product in great detail at CES 2017 (“Artificial Intelligence for Autonomous Driving: State of Technology, Breakthroughs, and Why Alliances are Valuable”).
Many developers in the self-driving car sector hide their technology from the public. Mobileye, a company that specializes in optimized sensing platforms for driverless vehicles, has taken a completely different approach to such traditions by revealing various components of its product in great detail at CES 2017 ("Artificial Intelligence for Autonomous Driving: State of Technology, Breakthroughs, and Why Alliances are Valuable").
At the start of the one-hour press conference, Prof. Amnon Shashua (CTO and Co-founder of the company) elaborated on the three pillars of autonomous driving:
· Sensing: From forward-sensing capabilities to 360-degree awareness; environmental models of one's surroundings
· Mapping: A combination of HD maps and localized maps at 10cm accuracy; main issue includes deployment of updates
· Driving Policy: "Human-like" negotiation skills via algorithmic processes and machine learning (explained below)
According to the Mobileye CTO, currently, most self-driving systems are programmed to be defensive. This is great for maintaining safety, but ineffective for road engagement. For example, driverless cars that merge into traffic must be able to identify an opening on the road in a slightly aggressive (but safe) manner. Prof. Shashua highlighted the balance between "human negotiation" and "safety" is key to advancing driverless platforms. For autonomous driving, this level of negotiation must take place visually via "motion."
This level of negotiation on public roads can be made possible through machine learning. The old "rule-based approach" to assessing road scenarios is too simple for self-driving platforms, resulting in "simplistic driving polices." Mobileye, through machine learning, aims to streamline the observation and collection of data to solve driving policy issues in real-time.
With this in mind, the speaker clarified that SAE L3 autonomous platforms are only for highways, requires HD maps and features non-instantaneous Take Over Request (TOR) times. When it comes to sensing components for L3 self-driving vehicles, the vessels are equipped with a "cocoon" radar, front-facing LIDAR unit and tri-focal or wide-angle cameras.
Michael Cheng is a legal editor and technical writer with publications for Blackberry ISHN Magazine Houzz and Payment Week. He specializes in technology business and digesting hard data. Outside of work Michael likes to train for marathons spend time with his daughter and explore new places.
- July 25th, 2017 News of the Day: NIO Capital Leads $46M Round in Self-Driving Technology Firm Momenta, Toyota Developing Long Range Fast Charging EV Batteries
- June 23rd, 2017 News of the Day: Tesla might have its music streaming, Geely could move Lotus to China
- Terabyte Vehicle Storage Demands Will Increase Flash Memory Usage
- Tesla Has a Backup Plan if it Can’t Keep its Autonomous Promises
- LED Streetlamps in London Facilitates EV Charging
- General Motors Launches Dev Client for Developers to Test Apps in GM Cars
- Octo Telematics Launches its Next Generation IoT Platform
- August 11, 2017 News of the Day: Tesla Patents Pyrotechnic EV Battery Disconnect Device, Germany’s Autobahn to Get Overhead Power Lines to Charge Electric Trucks
- Volvo Unveils its New VNL Series Semi-Autonomous Trucks
- Nauto Closes on Another $159 Million in Funding From BMW, GM, and Others