Alibaba recently announced a last mile delivery robot called the G Plus. This is an important announcement. They are not the only company operating in this space, but they are the only company, according to Starship Vice President Henry Harris-Burland, that have moved beyond the pilot stage to actual commercial operations.
The robots can be used to deliver parcels, groceries and food on campus environments and within residential neighborhoods. One of the target markets is delivering food from restaurants or consumer goods from stores and specialized hubs to consumers and businesses within 15 to 60 minutes. The robots’ entire journey can be monitored on a smartphone. And the deliveries are inexpensive: $1.99 in the US or £1 in the UK!
The consumer downloads an app, makes the order, and then sticks a pin on the map where they want the goods delivered. The delivery point might be a home or business, or just somewhere the customer knows they will be. When the robot arrives, the customer uses their smartphone to open the bot via the app. In other words, there is a layer of security, as the robot is locked at all times and can only be opened by the recipient. One customer is public – the Co-op, grocery store in the UK – has partnered with Starship to sell their groceries.
Mr. Harris-Burland says they have two business models: “We are a delivery service, we run the robots and take delivery fees. Secondly, we work with partners in a variety of industries who pay us a fee in a Robot-as-a-Service (RaaS) model.” When I asked what the charges were for RaaS, Mr. Harris-Burland explained that it depended on the number of robots but that it was “no more expensive than a human alternative.”
The way the delivery service works is that the customer goes to the Starship website and adds items to their basket. They pay. The customer then picks a delivery location by dropping a pin on the map. That delivery location could be their home, their office, or it might be a local park where they are having a picnic. Either a Starship associate or store associates pick the items, puts the items into the robot outside the store, and then the delivery robot drives autonomously on sidewalks to the delivery location. When necessary, the robots cross roads at intersections.
The bots can carry 20 pounds in weight or the equivalent of three shopping bags in volume. The bot is not refrigerated. Products that need to stay cool or warm can be packed in special insulation bags. “We can keep (a product) hot or cool up to about an hour,” Mr. Harris-Burland stated. “For hot pizza, to ensure quality, we don’t take orders from consumers that are further than 45 minutes away, but we’re always improving our insulation methods.”
The robots’ sensors include radar, ultrasonic (a device that can measure the distance to an object by using sound waves), stereo vision, time of flight cameras (a time-of-flight camera is a range imaging camera system that resolves distance based on the known speed of light, measuring the time-of-flight of a light signal between the camera and the target). The different sensors allow the robots to operate during the day and at night and on both sunny and rainy days.
The navigation is based on Simultaneous Localization and Mapping (SLAM). With SLAM, the local delivery area is mapped prior to the robots being able to travel autonomously. Then robots use their sensors to triangulate themselves, based on their prior maps and what they see at that second in time. When moving objects, like a human, are detected, the robot stops or maneuvers around until the object clears their path. The robot moves at a maximum speed of 4 miles per hour, so humans are in no danger from the bots.
These delivery bots are not 100 percent autonomous. There are control towers, currently located in Estonia and Washington D.C., where the robot can call for help if needed. The humans can then take control of the robot, navigate around an obstacle or help it cross a street safely, and then let the robot take over again.
But the robots do not need help that often. “They cross the road autonomously the majority of the time,” according to Mr. Harris-Burland. “Our goal is to get to a point where one human controls around 100 bots.” Further, if the fastest route is likely to involve busy crossings that might require assistance from a human in a control tower, the bot might not have to go the fastest route. If the robot can meet the promised delivery time, speed is not of the essence.
When I heard about the delivery fee – $1.99 – I thout it was very likely that to make money Starship would have to scale. Why would scale allow Starship to become profitable, rather than just losing money ever faster? Mr. Harris-Burland agreed. A big network allows a bot to deliver from one store to a customer, then return to a different store owned by different retailer. Further, certain retail segments are “peaky.” For example, customers are more likely to order from restaurants during the lunch or dinner hour. Operating across different verticals will smooth demand. “The same bot that delivers shoes in the morning, could deliver food at lunch, and groceries in the evening. Our plan is to scale to 1,000 bots within a year.”