New AI Technology in Amazon’s Delivery Vans Makes Packages Obvious, Saving Drivers Time and Effort

Author: Ivy Levine

By early 2025, Amazon will deploy 1,000 electric delivery vans equipped with Visual-Aided Package Retrieval (VAPR).

Although finding packages at each stop takes only a few minutes, these times add up when delivering to hundreds of Amazon customers daily. With the new AI-driven solution VAPR, delivery drivers no longer need to organize parcels, read labels, or manually check critical identifiers such as customer names or addresses for every stop. They simply look for VAPR’s green light, pick up the package, and continue with deliveries.

The concept of VAPR originated in early 2020 when an internal team at Amazon Transport envisioned using technology as an invisible force to assist drivers. “We had to consider unique factors in the delivery experience, such as lighting and space constraints within the van,” said John Colucci, Product Manager at Amazon Transport. After years of refinement based on pilot driver feedback, VAPR will be rolled out across 1,000 electric delivery vans manufactured by Rivian by early 2025.

Locating the Right Package in Seconds

Once the van arrives at a delivery location, VAPR automatically projects a green “O” on the package to be delivered and a red “X” on other packages. Through audio and visual cues, VAPR confirms the correct parcel before the driver enters the cargo area. This technology also eliminates the need for drivers to use mobile devices during the process.

“Previously, it took me about 2 to 5 minutes to unload a box and organize parcels for the next stop. Now with VAPR, this step takes only around one minute,” said Bobby Garcia, a driver from Bloomfield Logistics who tested VAPR in the North Boston area of Massachusetts. “This has made my job much easier.”

AI-Driven Assistance for Drivers

VAPR employs Amazon Robotic Identification (AR-ID), a form of computer vision originally developed for fulfillment centers to automatically recognize items during inventory storage or picking operations. AR-ID replaces the need for manual scanning. It can locate and interpret multiple barcodes in real-time within processing environments. VAPR builds on this by training machine learning models to identify different labels and packages under various lighting conditions and package characteristics. The technology is optimized for the van environment, featuring specially designed automotive lamp projectors and cameras fully integrated into the van’s delivery route navigation system. VAPR also incorporates multiple AWS technologies such as SageMaker and IoT Greengrass on the road.

Effort and Time Reduction in Package Retrieval

VAPR streamlines the delivery process, allowing drivers to focus more on providing a superior customer experience and delivering accurate packages. “VAPR always finds the right package,” said Stephen Harry, owner of Delivery Service Partner (DSP) at Bloomfield Logistics.

The Amazon Transport team spent hundreds of hours in the field with DSPs and their drivers testing hypotheses to ensure these new experiences deliver value. Early tests have shown a 67% reduction in physical and mental strain on drivers, saving over 30 minutes per route.

With more than 390,000 DSP drivers and over 100,000 Amazon vans delivering millions of packages daily worldwide, VAPR can save significant time and effort for drivers at scale. We are excited about continuously improving the driver experience through technological solutions and innovations that will be applied across Amazon’s delivery network.

Source website: