App-based AI Smartly Navigates Away from Failed Deliveries

App-based AI Smartly Navigates Away from Failed Deliveries

With the surge of online-based purchases due to COVID-19 lockdown restriction since last year, more and more items are being delivered rather than bought in-store.

Statistically, this also means an increase in failed deliveries, with delivery personnel accumulating lost time and money over absent recipients.

To combat this, JDSC, Tokyo University, Sagawa Express, GDBL, and officials at Yokosuka city have all joined together to complete a proof-of-concept test using a delivery rerouting system.

For the app’s main efficiency results, it has shown a marked 20% decrease in the overall rate of failed deliveries during the entire experiment period, which is assumed to have been conducted during October to December 2020.

Surprisingly, its design has already been in development long before the viral outbreak even began.

Prototekton: “Zero Redelivery AI Project”

The year was 2020. The Japan Data Science Consortium (JDSC), amidst the height of COVID-19’s first few months of economic onslaught around the world, unveiled a peculiar smartphone-based technology.

This technology, powered mainly by advanced AI, aims to recalculate routes of delivery vehicles. As you may have guessed already from the article title, the main element of consideration would be whether the package will be properly received at the destination.

The project was tentatively named Prototekton, which awfully sounds like a dollar store version of some 1970’s-esque super robot character.

It takes the form of a smartphone app, which overlays a map that indicates delivery routes, the points where the package may not be received, and plots the optimal route that would complete the deliveries in the shortest amount of time.

Now, as to whether looking at months or years of electric consumption data would actually yield reliable accuracy for the AI, we don’t know. For example, if you’re very thrifty in spending electric power for the last few months, the AI might deem that you’re not really chilling at your home, and skip you and your delivery items. Though, based on the number of tests done prior, it seemed to have been working well enough to reach a wider, city-scale trial implementation.

As for its development, the very first working iteration of the “Prototekton” was implemented two years before its unveiling. Around September to October of 2018, a very small-scale proof-of-concept test was conducted within the Tokyo University campus, bringing positive results to the idea and legitimizing its efficiency.

Furthermore, in September 2019, the very same prototype system was used by Sagawa Express in a computer-simulated delivery trial run. The result did not seem to be as promising as the first. But it has been reported that its system development system was extended to at least three more groups by the end of October 2019.

Initial Urban Field Test Results

The next field test was supposed to be conducted sometime during the last quarter of 2020. The objective was to use five groups to scour an intended “B route” within Yokosuka city. Preparations were announced as early as July. However, no updates were officially reported up until the recent announcement this week.

The official timeline of the latest proof-of-concept trial was supposedly during October to December 2020.

Using significantly updated and integrated residential data around Yokosuka, the app was provided to drivers of different positions and experience levels (regular delivery personnel, substitute employees, fresh recruits, etc.). Based on user experience alone, the researchers found out that there is no significant difference in the app’s usage efficiency between delivery personnel of varied experience levels.

This may not seem much, but when factored in with the people who did a follow-up call to the delivery company on the same day they were absent, this becomes quite significant.

Long story short, this resulting data pattern lets the AI system’s priority shift from the previously simpler “optimizing delivery routes” to the relatively more specific “creation of failed-delivery-avoidance routes.” Presumably, a sufficiently advanced version of this system would leave same-day follow-up failed deliveries as the only remaining record of failed deliveries for a company implementing the system.

Japan’s Re-Delivery Challenges

In the last few decades, even before the current global pandemic hit, Japan has already experienced exponential growth in its electronic commerce industries. Related business got bigger as a result, though the number of services available online has also been continuously increasing. This directly correlates to a massive increase in carbon emissions and demand for more drivers to serve the every-booming industry.

In fact, according to the Ministry of Land, Infrastructure, Transport and Tourism, just in October 2020 the country’s delivery industries has experienced a total of 11.4% redelivery rates. This is roughly equal to about 300,000 deliveries that required second or multiple revisits by delivery personnel, incurring somewhat considerable losses in time any money.

Most interestingly, JDSC also shows that within all the years that redelivery rates were recorded, the typical average was usually about 20 percent. With the average 25 percent loss on covered distance and 90,000 delivery personnel nationwide, the organization has calculated this to an annual 200 billion yen (1.8 billion USD) of net loss every year.

As some of you may have imagined already, an interesting question to this, is how does this technology play with the current driverless technologies we have? Surely the automated navigation systems that are specifically honed for delivery services at the moment could be integrated in some manner to offset redelivery losses right?

Well, here’s to hoping that a workable Level 5 system is available for us to find out soon enough.

Sources:

Leave a Reply

Your email address will not be published. Required fields are marked *

Previous Post
Cheap 240Hz Monitors

Best 19 Cheap 240Hz Monitors – Buying Guide (2021)

Next Post
Best Thunderbolt 3 Raid Storage Systems

7 Best Thunderbolt 3 Raid Storage Systems – Buying Guide

Related Posts