Delivery Tech WatchArticles

Automating Last Mile Delivery #1

Where are robot and drone delivery companies operating? What are they promising, and can they deliver on those promises? Are current business models sustainable, or will these companies have to explore different avenues to become profitable? These questions and more are what I am to research and hope to be able to answer over time.

Welcome to the first entry of my newsletter tracking robot and drone delivery operations with a focus on the US. I'm interested in seeing how these new industries develop, and I haven't seen as many blogs or websites dedicated to them as compared to autonomous vehicles. I'm starting this newsletter as well as this website to track their expansions (and pivots).

Featured: Improving Situational Awareness

Avride currently has a pipeline in place using a vision-language model (VLM) to select footage from situations of interest that their delivery robots encounter for further training and analysis. Now, they are adapting that to provide their robots with greater awareness to better deal with complex situations such as emergency responses and construction zones. When a robot encounters something it is unsure about how to handle, the VLM can analyze and escalate to a human remote assistant as necessary. You can read more about how it works here.

In fact, there recently was an event where DoorDash's delivery robot, Dot, drove into a SWAT scene in Chandler, Arizona. A DoorDash technician had to come pick up the robot and take it away in a box truck, according to 12news.com. They don't describe to what degree it disrupted the crime scene, but it is undesirable nonetheless and presumably whoever ordered delivery did not end up receiving their order.

It is not an issue limited to delivery robots and could be considered a problem generally with autonomous systems that operate in public areas. This month, the NHTSA issued a letter addressing autonomous vehicle (AV) companies about the trend of driverless AVs interfering with emergency responders by blocking paths or driving into scenes, giving the companies until the end of this month to describe how they will solve this behavior.

Of course, AVs are larger than delivery robots and also carry passengers instead of items, so there is the potential for real danger when they fail to respond to instructions or recognize an emergency situation. But if even the AV companies which have more to gain from solving this problem are struggling, then perhaps it implies that despite leaps in artificial intelligence in recent years, we are still far from imbuing machines with any kind of common sense about the physical world that would enable them to independently deal with such situations the way most humans do.

Delivery robot companies have plenty to gain, too — it affects how quickly a customer can get their order compared to a human courier and whether a pedestrian can share the sidewalk or has to go around a stuck robot. It sounds like Avride is taking a step in the right direction and having these robots in my city, I'm curious to see if they observably get any better at navigating!

Top Picks

Inside the U.S. race for drone delivery dominance (Reuters)

US drone regulations aren't well-equipped to deal with autonomous delivery drones, and the steps drone companies have to go through illustrate that. For example, current regulations require visual line of sight with a drone at all times when it is flying. This means posting observers on the ground at regular intervals along the flight path (remote assistance doesn't count towards the requirement) which has a significant impact on the size of the service area and the operational costs. Companies are able to apply for exemptions to this, but the process can take a while. Now, new regulations have been proposed which would better fit the operational logistics of autonomous drone delivery.

This robotics CEO wants to automate the work that makes people quit (Business Insider)

This interview with Robot.com's (Kiwibot) CEO, Felipe Chavez, focuses on the company's expansion beyond delivery robots to warehouse and humanoid robots. An interesting anecdote Felipe brings up is how adding robot delivery to a restaurant automated one task but exacerbated another one — putting the food deliveries into the robots — which led to some businesses having to retrain or even hire more staff. This shows the math isn't as simple as replacing humans altogether, and brings up the question of how much “boring work” robots eliminate compared to how much they might create.

Safety at Serve, from Principle to Practice (LinkedIn)

This article by Jake Streeter, Serve Robotic's Head of Safety, gives insights into the engineering decisions behind making their delivery robots safer and pedestrian interactions smoother. Design factors include the kinetic energy of the robot, what temperatures the exterior can reach, and matching how pedestrians move and infer direction to make sharing the sidewalk easier. It's an interesting read for anyone who wants to learn more about what goes into designing a delivery robot.

That's all from me — thanks for reading! Stay awesome and see you next time!

This newsletter is also published on my Medium and Substack accounts. Subscribe there to get the latest edition in right your inbox.