Robot Dogs at Your Door: How Boston Dynamics Is Testing Spot for Last-Mile Delivery
7/21/2026
Spot, Boston Dynamics' four-legged robotic workhorse, has already proven its value in environments most robots avoid — crawling through industrial plants to flag equipment anomalies, navigating the uneven cobblestones of Pompeii, and conducting hazardous-site surveys. Now, the company is turning its attention to one of the most persistent headaches in modern commerce: getting a parcel from a delivery vehicle to your front door.
The Last-Mile Problem, Explained
"Last-mile delivery" refers to the final leg of a shipment's journey — from a local distribution hub or vehicle to the recipient's address. It sounds simple, but it's consistently the most expensive and labor-intensive part of the entire supply chain. Uneven terrain, steps, locked gates, narrow pathways, and the sheer volume of stops make this segment stubbornly resistant to automation. Conveyor belts and warehouse robots handle bulk sorting brilliantly, but the moment a package needs to travel across a front yard, the logistics industry still largely depends on human hands and feet.
That's exactly the gap Boston Dynamics is trying to close with Spot.
What the New Accessory Does
The company is developing a conveyor belt attachment that mounts onto Spot, enabling it to receive packages directly from a delivery vehicle and transport them autonomously to a designated drop-off point — typically a customer's doorstep. Rather than a delivery driver walking each parcel to the door, Spot handles that final stretch independently, then returns to the vehicle for the next package.
The key word here is autonomously. Spot navigates using a combination of onboard sensors, cameras, and real-time environment mapping. It can handle irregular surfaces, steps, and obstacles that would stop a wheeled robot cold. The conveyor mechanism streamlines the hand-off from vehicle to robot, reducing the need for human intervention at each stop.
The stated goal is reducing the physical burden on delivery workers — a role that carries a notably high rate of musculoskeletal injury due to constant lifting, bending, and walking across varied terrain under time pressure.
Why Quadrupeds Make Sense Here
Wheeled delivery robots work well on flat urban sidewalks, and aerial drones excel at covering distance quickly. But quadrupeds occupy a useful middle ground: they move at a reasonable pace, carry meaningful payloads, and — critically — they can navigate the same physical environments humans do. Steps, grass, gravel, and sloped driveways are manageable rather than impassable.
This is precisely the design philosophy behind a growing class of industrial quadrupeds. The Unitree B2, for example, is built with a walking payload capacity exceeding 40 kg, IP67 weather sealing, and the ability to traverse rough terrain at speed — characteristics that align closely with the demands of real-world delivery environments. For developers and researchers who want to explore quadruped locomotion and logistics applications at a lower barrier to entry, platforms like the Unitree Go2 offer 4D LiDAR perception and an open SDK, making them accessible tools for prototyping autonomous navigation workflows.
The Bigger Picture: Edge AI Is the Enabler
None of this works without capable onboard computing. Autonomous navigation in unstructured environments requires a robot to perceive its surroundings, build a map, plan a path, and respond to dynamic obstacles — all in real time, without offloading heavy computation to a remote server. The trend toward powerful edge AI hardware is what makes this practically viable.
Developer platforms like the NVIDIA Jetson AGX Orin 64GB illustrate how much compute can now be packed into an embedded form factor — delivering the kind of inference performance that once required a full server rack, now running on a device that can ride aboard a mobile robot. As these platforms become more accessible, the gap between research prototype and commercial deployment continues to narrow.
Challenges Still Ahead
Boston Dynamics is clear that this is still in the testing phase, and there are real hurdles to clear before robot dogs become a routine sight on residential streets.
Regulation and liability are significant open questions. Who bears responsibility if Spot damages a property, startles a pet, or fails to deliver in adverse weather? Frameworks for autonomous ground robots operating on private property remain underdeveloped in most jurisdictions.
Public acceptance is another factor. Surveys consistently show that people have mixed feelings about robots navigating their neighborhoods — a very different context from a robot working inside a factory.
Edge cases abound. A front gate left open, a child's bicycle blocking the path, or a sudden downpour all represent scenarios that require robust real-world testing before any company can responsibly scale.
What This Tells Us About Robotics in 2025
The Boston Dynamics delivery experiment is a useful signal about where the industry is heading. The emphasis isn't on replacing the delivery driver wholesale — it's on redistributing the workload, specifically the physically demanding repetition of the final walk from vehicle to door. That framing matters, both for public reception and for the practical reality that human judgment remains essential for handling exceptions.
What's becoming clear is that last-mile logistics is emerging as one of the most active proving grounds for autonomous robotics outside of controlled warehouse environments. The companies — and the hardware platforms — that crack reliable, safe, outdoor autonomy at scale will have solved something genuinely hard.
Spot may or may not end up delivering your next package. But the engineering being developed to get it to your doorstep will shape how robots navigate the real world for years to come.
Interested in exploring quadruped robotics or edge AI platforms for your own logistics, inspection, or research applications? Get in touch with our team to discuss which solutions fit your use case.
References
This article was drafted with AI assistance and reviewed before publishing.