Autonomous Delivery Robots: The Current State of Last-Mile Automation

Autonomous delivery robots are no longer prototypes — they are operating in dozens of cities and university campuses across the US and Europe. These small, wheeled robots navigate sidewalks and crosswalks to deliver food, groceries, and packages. Here is how they work and what the current experience is like.

How Delivery Robots Work

Most delivery robots are cooler-sized, six-wheeled vehicles that travel on sidewalks at walking speed (3-4 mph). They use cameras, LiDAR, ultrasonic sensors, and GPS to navigate. Machine learning algorithms identify obstacles, traffic signals, crosswalks, and pedestrians. A remote operator can take over if the robot encounters a situation it cannot handle.

You place an order through a delivery app, the restaurant or store loads the order into the robot's compartment, and the robot navigates to your location. You unlock the compartment with a code from the app. The entire process is contactless.

Major Players

Starship Technologies has deployed over 6,000 robots across university campuses and cities, completing millions of deliveries. Their robots are the most commonly encountered, operating at over 25 universities and in several cities including San Francisco, Los Angeles, and cities in the UK.

Serve Robotics (spun out of Uber) operates in Los Angeles, delivering for Uber Eats. Their robots are slightly larger than Starship's and handle a wider range of restaurant deliveries. Kiwibot operates on college campuses with smaller, lighter robots focused on food delivery.

Amazon Scout tested delivery robots in several neighborhoods before pausing the program to focus on drone delivery. Other companies including FedEx and Nuro (which operates larger road-going vehicles) are developing autonomous delivery at various scales.

The User Experience

Ordering from a robot is straightforward — you select robot delivery in the app (where available), and the process is similar to any delivery. Delivery times are typically 15-30 minutes for campus deliveries. The novelty factor is high — watching a cooler-sized robot trundle up to your door is inherently entertaining.

The trade-offs are limited capacity (one order at a time, no stacking), slower speed than car delivery, weather sensitivity (most robots struggle in heavy snow), and limited operating areas. Robot delivery works best for short distances in pedestrian-friendly environments.

Economics

Delivery robots aim to solve the economics of last-mile delivery. Human delivery is expensive — paying a driver $15-20/hour limits how cheaply delivery can be offered. A robot's operational cost is primarily electricity and maintenance, making deliveries under $2 economically viable at scale.

This cost advantage is why universities and dense urban areas are the first deployment zones. Short distances, flat terrain, and high order density maximize robot utilization and minimize delivery time.

Regulation and Challenges

Sidewalk delivery robots operate under a patchwork of local regulations. Some cities welcome them, others have banned or restricted them. Pedestrian safety concerns, accessibility advocates worried about sidewalk obstruction, and general resistance to automation have slowed deployment in some areas.

Technical challenges include navigating construction zones, handling aggressive pedestrians (some people kick or vandalize robots), crossing busy streets, and operating in weather extremes. Remote operator intervention rates are decreasing as AI improves but remain necessary.

The Future

Delivery robots will expand to more cities and campuses as the technology matures and costs decrease. Integration with smart home systems could enable unattended delivery — the robot communicates with your smart lock to leave packages inside. Larger road-going autonomous vehicles (like Nuro's) will handle bigger deliveries.

The long-term vision is a mesh of autonomous delivery options — sidewalk robots for meals and small items, road vehicles for groceries, and drones for urgent lightweight packages. Each mode optimized for its ideal use case.

---

As an Amazon Associate, BestElectronicsReviewed earns from qualifying purchases.