Local goods delivery is an area that has exploded in popularity over the past year of the pandemic and going forward, it is expected to be one of the biggest applications of automated vehicles (AV). One of the many AV companies that has been focused on this particular segment of the market is Burlingame, Calif.-based Udelv. Udelv has now decided to discontinue the development of its own automated driving system (ADS) and adopt the Mobileye Drive system for its purpose-built Transporter delivery vehicle.
Udelv has until now been developing an ADS based on the Baidu Apollo open-source platform. That ADS has been installed on the Ford Transit Connect compact cargo van. Udelv will now begin the transition to a purpose built electric delivery vehicle with the first prototypes of the Transporter expected to hit the road in 2022 ahead of first customer deliveries in 2023.
Like most purpose-built AVs, the Transporter looks like a big shoe box on wheels with sliding doors on either side to provide access to the cargo storage. The Transporter is being built on an electric skateboard platform, but Udelv CEO Daniel Laury isn’t ready to reveal the source of that platform just yet. There are however, numerous options available from companies including Rivian, Bollinger and Ree.
Laury did acknowledge that the range is expected to be about 250-300 miles which should allow it to operate most of a normal working shift without needing a recharge. Since there is no driver compartment, it won’t require a climate control system that robs from the range. The sliding doors on the side can automatically open up to individual compartments within the vehicle for each delivery.
While Udelv is using Mobileye Drive, it has developed its own teleoperation system which is claimed to have a latency of only 105 milliseconds. Early on, this is expected to be utilized for navigating complex parking lots and loading docks at the back of stores and warehouses that have not yet been mapped.
Once on the road, the Transporter will take advantage of Mobileye’s crowd-sourced Roadbook map platform to navigate and also help it make path planning decisions. Roadbook, which is based on the Road Experience Management system uses the camera sensors in vehicles with driver assistance systems running on the Mobileye EyeQ4 chip to capture road information.
In addition to the road configuration and static features adjacent to the road that can be used to localize the vehicle, Mobileye also captures data about the typical speeds that human drivers go on particular roads and where they actually stop when making turns. Since the official stopping points at intersections are often set back for safety reasons, drivers will pull ahead and stop at a position where they can get a clear view of oncoming traffic before making a maneuver like a right turn at a red light or stop sign. Since the sensors on an AV are limited to line of sight, just like a human driver, learning from humans about where to stop for a clear view can be a significant aid to ADS.
There are already more than 1 million vehicles on the road on road from BMW, Nissan and Volkswagen with EyeQ4 that have collected more than 700 million km of road data with about 8 million km being added daily. By 2024, Mobileye expects to be updating or collecting 1 billion km per day. All of the data is anonymized and aggregated on servers for verification and then distributed back to users.
The Drive ADS will use 13 cameras, 6 radars, 3 long-range lidar, 6 short-range lidar and 2 EyeQ5 processors in the same basic configuration as the robotaxi applications that Mobileye is developing. This will provide what Mobileye calls True Redundancy. Only the camera data is to build the environment model of what is around the vehicle to classify other road users. That model is then used to make decisions about the path that the vehicle should take. The radar and lidar are fused together and run through an independent environment model that is passed through Mobileye’s Responsibility Sensitive Safety (RSS) model which is designed to ensure that the ADS never issues a command that will cause or contribute to a crash. RSS essentially acts as a sanity check on the camera-based ADS.
Mobileye recently announced that it was developing its own imaging radar and lidar sensors with parent company Intel . However, those sensors won’t be ready for initial production of the Transporter in 2023 and Mobileye isn’t ready to reveal the sensor providers yet. Mobileye is currently using Luminar lidar sensors on its robotaxi development fleet. Mobileye plans the first commercial launch of robotaxis in Israel in early 2022.
Udelv has received the first pre-orders for the Transporter from commercial fleet leasing and management company Donlen. Donlen plans to take 1,000 of the automated delivery vehicles and will provide delivery as a service through a monthly subscription. Udelv is targeting to build 35,000 examples of the Transporter between 2023 and 2028.