Autonomous shuttle company EasyMile and sidewalk robot delivery startup Coco are integrating DriveU.auto's teleoperation and connectivity platform into their operations, DriveU.auto, an Israeli startup that came out of stealth last year, they announced at CES.
The autonomous vehicle industry is still a ways off from being able to commercialize full self-driving technology, despite the promises certain companies have made or the way they choose to name their advanced driver assistance systems. Indeed most countries still require a human to be in the loop during autonomous operations for safety purposes. To get to market faster and increase public acceptance of driverless vehicles, AV companies are turning to teleoperations, where a remote driver can swoop in to pilot the vehicles in the event of an emergency, anomaly or safety incident.
"Think of the site of an accident where there are multiple police officers gesturing at traffic to drive around," Alon Podhurst, CEO at DriveU.auto, told TechCrunch. "The AI of the vehicle will in all likelihood ask for assistance to interpret these gestures and vocal commands. The remote operator needs to see a real-time view of the world around the vehicle she is assisting -- a robot or an autonomous vehicle...We want to stream the feed from the vehicle sensors to the remote operator’s location. This must be done in a manner ensuring reliable, high-quality and low-latency connectivity so that the decisions being made by the remote operator are based on the actual real-world situation around the vehicle. This is done over cellular networks."
Much of the success of teleoperations relies on high-performance connectivity in order to transfer video, audio and other sensor data. DriveU.auto's connectivity platform aims to ensure stable network connections and avoid any latency, "dark spots" or drops in connection that could prevent a teleoperator intervening to help an autonomous vehicle.
"No single cellular network -- not even 5G -- can guarantee the performance levels needed for reliable remote operation," said Podhurst. "Remember, the task is to transmit multiple feeds, as there are multiple cameras on the vehicle, of high-definition video over a constrained cellular network, from a moving vehicle. Bottom line -- one network is not enough."
DriveU.auto's tech is already live on a fleet of EasyMile's EZ10 autonomous shuttles that are serving a medical complex in France, and the company is in the process of integrating into the entire EasyMile fleet, according to Podhurst.
“As we continue to deploy more and more use cases for autonomous vehicles, we expect remote supervision to be a key component of our solution," said Benoit Perrin, managing director of EasyMile, in a statement.
The startup's connectivity solution is also already in Coco's fleet of about 100 Coco 0 units, its proof-of-concept pilot vehicle. The plan is to integrate the platform into the new shipment of 1,000 Coco 1 robots, the hardware base of which is being built by Segway and will be deployed in Los Angeles and two other U.S. cities during Q1 2022, according to Coco.
Aside from EasyMile and Coco, the startup says it is already operational on robotaxis, autonomous trucks, other delivery robots and special use case AVs, all partnerships that are still under non-disclosure agreements, but Podhurst says he hopes to go public in the coming weeks. The company recently announced an 18-month partnership with Denso, a Japanese auto parts manufacturer.
DriveU.auto typically provides customers with a software development kit that is integrated into the vehicles' computers. Customers operate their own vehicles, including the teleoperation, relying on the vehicle's existing sensor suite and other hardware components. Podhurst says this software-only approach is a key component in the company's market traction because it makes for faster integrations.
The software-based connectivity platform works by fusing three technologies: dynamic video encoding, low-latency algorithms and cellular bonding. The fused data package is then sent over multiple cellular networks based on the performance of the networks at the time of the transmission. The data is then reconstructed as video frames once it reaches the remote operator side. To break it down further, the platform is essentially made up of software modules that are integrated into the vehicle's system and that interoperate with both a cloud-based software component and a module that's been integrated into the remote operator's computer.
“Integrating a superior connectivity solution on delivery robots dictates extremely demanding power and compute parameters,” said Sahil Sharma, COO of Coco, in a statement. “Having evaluated the industry leaders in this space, we found DriveU’s solution to be the strongest match for our growth plans and aggressive delivery schedules.”