Computer Vision
7 min read
Delivery Robots: Solving the Last-Mile Delivery Problem
Written by
Chinar Movsisyan

Over the last five years, residents of cities around the world have welcomed a new type of vehicle to their streets and sidewalks. Delivery robots, or autonomous delivery vehicles, are quickly being adopted for what is known as last mile delivery; the process of getting a package to its final destination during the last leg of its journey. By utilizing recent advancements in machine learning, computer vision, and computer vision robotics technologies, autonomous delivery vehicles are tackling this specific problem. They are typically used for delivering small packages, food, and other goods to homes and businesses. According to Allied Market Research, the global robot delivery market is estimated to grow nearly 10x during this decade (2020-2030).  

What is Last Mile Delivery?

The end-to-end supply chain process, beginning with a business receiving an order and ending with the customer receiving the item they purchased, can involve several steps in between, such as the item being shipped from a warehouse the business operates in another region, to one closer to the customer’s address. The final step, getting the order to the customer is referred to as the last mile. This is the most important step, as if it is not executed successfully, the order will be unfulfilled.

Last-mile delivery is typically the most expensive part of the supply chain process, largely due to increased labor costs that are associated with it. A driver delivering packages will often have to navigate through densely populated urban areas with heavy traffic, which can slow down delivery trucks and increase fuel consumption. Additionally, last mile deliveries are typically made to individual residential addresses, which can be difficult to access and plan an efficient route for, leading to longer delivery times and higher labor costs. Furthermore, the cost of maintaining and operating a fleet of delivery vehicles can be high, and these costs are passed on to consumers in the form of higher delivery fees.

How Delivery Robots Solve the Last Mile Problem

The demand for last-mile delivery has gone up significantly in recent years as e-commerce has become more popular. With more people shopping online, there is a greater need for efficient and cost-effective delivery services that can get packages to customers' homes quickly and reliably. However, as the demand for last-mile delivery has increased, the capacity to meet that demand has not kept pace. To solve this problem, many companies are turning to delivery robots, which are currently being used in various forms and by various companies.

Some companies are using ground-based robots to deliver packages to customers' homes in residential areas. These robots can navigate sidewalks and streets, and they can carry a variety of different types of packages. Others, such as Amazon, are experimenting with using drones to make last-mile deliveries. Overall, the use of delivery robots is still in its early stage, but with the increasing demand for last-mile delivery, it is expected that more companies will adopt this technology in the future.

The advantages that delivery robots provide are plentiful. The primary benefit is that they don’t require human labor to conduct the last mile delivery. This significantly cuts the costs associated with last-mile delivery for businesses. Delivery robots also enable faster delivery. If a business has access to a fleet of robots, there is no longer a need to wait for a delivery person to become available to deliver the goods. It also frees up workers to work on more important tasks.

Computer Vision

Delivery robots rely on a multitude of sensors and cameras to make sense of the world around them. Much like autonomous vehicles, delivery robots also rely on computer vision to detect objects in the environment they operate in. In order to safely conduct its function, the delivery robot must be able to detect humans and other obstacles that may be in their path as they attempt to reach their destination.

When comparing delivery robots with self-driving cars (SDCs), there are a few noticeable differences in the computer vision problems that delivery robots have to solve. Delivery robots often operate on sidewalks, whereas SDCs operate on roads. The environment provided by a road is largely structured. Cars drive within lanes and, for the most part, drive straight ahead. Sidewalks can be more dynamic. People do not walk on sidewalks in a structured manner. Someone may be jogging, whereas another person may be walking slowly with a stroller. There are also other factors such as objects or animals that could be found on a sidewalk.

Neural networks work extremely well when the data that they are working with is within their data distribution. In other words, if the model has significant exposure to a scenario in its training data, it will be able to accurately perform its function when encountering it in the real world. As with any computer vision problem, it is important to continuously monitor the performance of the model in the production environment to detect outliers. Outliers are samples of data that fall outside of, or far from, the model’s data distribution. By catching these outliers using model observability tools, those new samples of data can be added to the training dataset of the model in order to improve the accuracy of the model.

Examples on the Market

One example of last-mile delivery robots on the market today is Tiny Mile. These small pink robots began operating in Charlotte, North Carolina last year, and later expanded to Miami, Florida. Small businesses, such as restaurants or shops, can use Tiny Mile to deliver orders to their customers. They are also being utilized as a courier service for documents, or even sending gifts to friends and family that live locally.

In the European market, German company THEO has launched their own delivery robots that are optimized for operating in bike lanes, however are agile enough to be able to operate on street as well. They are larger than the average delivery robot and can carry payloads up to 100 kg. THEO delivery robots are also emission free.

Stay up to date !

Subscribe to our newsletter to get inbox notifications.
Sign up to our newsletter ↓
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.