Autonomous Last-Mile Delivery: Robots, Drones, and the Final 100 Meters
Last-mile delivery — the journey from a local hub to the customer’s door — accounts for over 50% of total shipping costs. It is the most labor-intensive, least efficient, and fastest-growing segment of the logistics chain. As e-commerce volumes continue to climb and consumer expectations for speed tighten, the economics of human-driven delivery are becoming increasingly strained.
At Güil Mobility Ventures, we see autonomous last-mile delivery as one of the most practical near-term applications of robotics and autonomy — precisely because the operating environment is more constrained and predictable than the open road.
Sidewalk Robots: Slow, Steady, and Scaling
Sidewalk delivery robots — small, low-speed autonomous vehicles that navigate pedestrian infrastructure — have moved from novelty to genuine utility in a growing number of markets. Starship Technologies has completed over 6 million autonomous deliveries across university campuses and suburban neighborhoods, operating in conditions that would be considered edge cases for full-size autonomous vehicles: rain, snow, nighttime, and crowded pedestrian zones.
The economics are straightforward. A Starship robot completes a delivery for $1.50–$2.00, compared to $8–$12 for a human driver in most US markets. The robots operate 16–20 hours per day, don’t require benefits, and improve with every delivery as the fleet’s mapping and navigation data grows.
Coco Robotics and Serve Robotics are pursuing similar models with slightly larger vehicles designed for restaurant and grocery delivery in urban environments. The common thread is a deliberate focus on low-speed, geo-fenced operations — environments where the safety case is manageable and regulatory approval is achievable.
Aerial Delivery: The Drone Promise
Drone delivery has been five years away for the last ten years — but the regulatory and operational picture is finally shifting. Amazon’s Prime Air and Wing (Alphabet) have received FAA Part 135 certificates for commercial drone delivery operations. Zipline, which built its reputation delivering medical supplies in Rwanda, is now operating in the US with a system capable of delivering packages weighing up to 3.5 kilograms within a 16-kilometer radius.
The use cases where drones genuinely outperform ground alternatives are specific: low-weight, time-sensitive deliveries to locations with limited road access or high traffic congestion. Pharmaceuticals, medical samples, and urgent spare parts fit this profile. The vision of drones delivering every Amazon package is neither technically feasible nor economically rational at current battery energy densities.
The Regulatory Patchwork
Autonomous delivery regulation varies dramatically by jurisdiction. Texas permits sidewalk robots statewide with minimal restrictions. San Francisco imposed a cap on the number of permitted robots after resident complaints about sidewalk congestion. The UK has no specific framework, leaving operators in legal limbo.
For drone operations, the FAA’s requirement for beyond-visual-line-of-sight (BVLOS) waivers remains the binding constraint on scaling. Each waiver is site-specific and takes months to obtain. A proposed rulemaking for standardized BVLOS operations would, if finalized, transform the scalability picture — but regulatory timelines are measured in years.
Where We Invest
Our investment thesis in autonomous last-mile delivery focuses on the enabling technology layers rather than the delivery operators themselves:
- Localization and mapping — centimeter-accurate maps of sidewalks, curb cuts, and building entrances that are essential for reliable autonomous navigation
- Fleet orchestration software — platforms that manage mixed fleets of robots and drones, optimizing routing, charging, and delivery sequencing across a geographic zone
- Sensor fusion at the edge — compact, low-power perception systems that enable small robots to navigate safely in unstructured environments
The last mile will be automated — but not uniformly and not overnight. The winners will be companies that match the right autonomy tool to the right delivery context, rather than pursuing a one-size-fits-all approach.
Resources
- McKinsey: Autonomous Delivery Robots — market sizing and deployment projections
- FAA UAS BVLOS ARC Final Report — regulatory roadmap for scaled drone operations
- Starship Technologies Impact Report — operational data from the world’s largest sidewalk robot fleet