Intel company Mobileye has added New York City to its growing global autonomous vehicle (AV) testing programme. The city boasts one of the most challenging driving environments in the world. With the tests, Mobileye aims to demonstrate the rich functionality of its AV technology as well as the rapid geographic and economic scalability of the Mobileye approach. Mobileye had previously tested its technology in Jerusalem, Munich and other cities. “Driving in complex urban areas like New York City is a critical step in testing the capabilities of an autonomous system and bringing the industry closer to market readiness,” comments Mobileye CEO Amnon Shashua,
As the video below shows, Mobileye’s camera-based subsystem handles busy streets filled with pedestrians, cyclists, aggressive road users, double-parked vehicles, construction sites, emergency vehicles, tunnels and bridges. The equipment is complemented by a second redundant subsystem based on radar and lidar sensors as part of Mobileye’s True Redundancy approach.
With permission to test self-driving vehicles in New York City, Mobileye is currently the only company to hold such a permit. The demanding test drives take place both during the day and at night. Seven factors are the focus:
- Pedestrians: Carelessly crossing the road is common in many cities. Due to the high number of pedestrians in New York, this behaviour poses an even higher risk. A self-driving car must make assumptions about the behaviour of these road users and incorporate them into its driving strategy. Humans do this instinctively, whereas machines must be programmed to do so.
- Driving behaviour: When roads are congested, drivers tend to become impatient and aggressive. Drivers in New York City – especially taxis and other commercial vehicles – are known to be much more energetic than in other cities.
- Traffic density and diversity of road users: Although the number of car owners in New York City is small compared to other large US cities, the diversity of road users is particularly high. Numerous taxis, limousines, buses, trucks, food trucks, emergency vehicles, bicycles, scooters, skateboards and even horse-drawn carriages make for a high traffic density.
- Second row parking: Which car is parked and which is not? The question is easier for a human to answer than for a machine. New York’s population density contributes to a high number of delivery vehicles that have to stop to unload. As a result, second-row parking is ubiquitous – a challenge for the electronics and algorithms of autonomous vehicles. The video shows how Mobileye’s vehicles take cues from other road users to decide when to take evasive action.
- Roadworks: As Mobileye’s continuously updated AV map also shows, NYC is one big construction site. While competitors either rely on their own test vehicles to create maps or invest millions of US dollars in special mapping vehicles, Mobileye receives data about blocked or blocked lanes from vehicles already on the roads. This data is also licensed back to city authorities.
- Tunnels and bridges: The island of Manhattan is connected to the surroundings via 15 tunnels and 21 bridges. Many of these are navigable via narrow lanes enclosed by bollards or pylons – the achilles’ heel of some autonomous vehicles. Despite this infrastructure and double-decker lanes, Mobileye’s crowdsourced mapping technology combined with recognition systems understands these obstacles.
- Lights and neon signs: Visual impacts and light pollution are challenges for an AV’s recognition system. Mobileye claims that its technologies overcome these thanks to fewer adjustments to the algorithm.