In 2018, a Tesla car in autopilot mode crashed into a highway divider in California, killing the driver and fueling doubts about the future of self-driving cars. The vehicle’s inability to identify the divider as a hazard revealed the limits of autonomous cars, especially their ability to recognize objects in the road – a key safety concern as more cars begin to incorporate automated driving features.
“Overreliance on Tesla’s ‘Autopilot’ and the operational design of Tesla’s ‘Autopilot’ have led to tragic consequences,” National Transportation Safety Board chairman Robert Sumwalt said after investigating this and other accidents. Vehicles belonging to Waymo and Uber, which recently sold its self-driving division, have been involved in similar accidents when cars’ sensing systems failed to detect other vehicles, pedestrians or objects in the road.
“The sensing system of cars is replacing the drivers’ eyes, and it needs to be perfect,” said Yuval Engelstein, a mobility research analyst at Start-Up Nation Central, which tracks tech companies in Israel. “Getting this perfect is essential.”
Israeli startup Arbe Robotics says its ultra-high-resolution radar system can provide this perfect vision for autonomous and semi-autonomous cars. The system recognizes objects, including people, up to 300 meters away, giving a driver – or an automatic braking system – ample time to stop. Unlike other advanced imaging and sensing systems, Arbe’s technology also works in low-visibility conditions caused by bad weather or glaring sun, and can detect cats, dogs and other small objects, as well as hidden and unseen dangers, such as a child walking between closely parked cars.
“It can really see and detect in a reliable way,” said Kobi Marenko, CEO of Arbe, whose systems are being developed in partnership with major global automotive manufacturers and will be available in some vehicles from 2022.
Traditional radar – which emits radio waves and uses a receiver to create images based on how those waves bounce off objects – has been around for decades, used mainly by the military and air traffic control. More recently, radar has also been deployed in consumer vehicles, working alongside cameras to assist features like warnings when cars swerve out of lanes.
But regular radar can only detect moving objects. It cannot detect elevation, like the height of bridges, and often cannot differentiate between objects close together, like a motorcycle next to a truck.
“This is the main weakness of today’s radar,” Marenko said.
Most systems for fully self-driving vehicles, as well as partially autonomous cars and driver-assist solutions, also rely on cameras, and sometimes LiDAR, which uses laser waves to detect objects. LiDAR creates higher-resolution images than radar and works in the dark, but is much more expensive.
Even these advanced systems are not perfect, Marenko said. “If there is sun shining directly on the camera, it’s a problem,” he noted. LiDAR can also fall short in snow, fog, or bright light.
Arbe’s technology improves radar performance, and eliminates the need for more expensive LiDAR. Arbe says it is the only company in the world that has developed such high-resolution radar, based on three computer chips: one for transmitting, one for receiving and one for processing images.
“It’s 50 times better than standard radar today,” Marenko said. “It’s a much safer solution.” The chips also contain a built-in solution to eliminate interference, another common challenge for radar, he said.
“I cannot elaborate about it a lot,” he said. “But it’s a unique approach, and today we are the only company that has a solution to this problem.”
The data-heavy system also works quickly because its processing chip uses algorithms to prioritize only the most important data coming from the receiving chip.
“We understand very fast, within nanoseconds, where there are objects, and do the ultra-high-resolution frame only on areas in the environment that have some information that is relevant,” Marenko said.
Importantly, it costs about the same as standard car radar systems, making it accessible. “It’s affordable technology here and now,” Marenko said. “You don’t have to wait 10 years until the price goes down.”
But the company noted some cameras will still be needed on cars because even high-resolution radar cannot detect color or read road signs. That could be achieved by working alongside camera-based solutions like the one developed by Israel-based Mobileye, now part of Intel.
Arbe started out by focusing on fully self-driving vehicles, but it has pivoted to supply the much larger market of vehicles that incorporate automated driver assistance systems (ADAS), like automatic emergency braking systems and features that keep cars from swerving out of their lanes. Better sensing technology is also key to this market, which is growing at a rate of more than 21% a year and expected to be worth about $189 billion by 2026, according to Allied Market Research.
“We decided to build a very scalable platform,” Marenko said. Demand is growing for such features in almost all new vehicles, while fully self-driving cars “will take a long time until they reach the mainstream,” he added.
The ADAS market has also received a boost in recent years from insurance discounts and new safety requirements, including European Union regulations that will make many advanced safety features mandatory by 2022 with the aim of saving thousands of lives.
“Our radar gives a full solution for this,” Marenko said, adding it can also be used in heavy machinery, drones and other specialized vehicles. “This is a golden age for automotive radar.”
For more information on Arbe Robotics, click HERE.