The Hidden Human Labor Behind Delivery Robots
Even as autonomous delivery robots become more common, a recent feature highlights the essential role of the human workforce that supports them. These teams handle maintenance, remote troubleshooting, and interventions, underscoring that full autonomy still requires significant human-in-the-loop infrastructure.
Companies like Serve Robotics and Starship Technologies operate their fleets at what is known as Level 4 autonomy; the robots handle most situations independently but rely on remote human operators to navigate complex "edge cases." Serve, which spun out of Uber and operates in Los Angeles, has a commercial agreement with Israeli firm Ottopia to use its teleoperation software, allowing remote staff to assist and control the robots when needed. This teleoperation requires a robust, low-latency connection, often using 5G and LTE, to transmit real-time video and sensor data to a human operator who may be miles away. These remote specialists take control during unpredictable events like navigating around construction sites, interacting with first responders, or crossing complex intersections—situations where the robot's onboard AI may lack the contextual understanding to proceed safely. Onboard, the robots are a fusion of advanced hardware, using a suite of cameras, radar, and LiDAR for perception. Much of this sensor data is processed at the edge using powerful AI platforms like the NVIDIA Jetson Orin, found in Serve's robots. For real-time motor control and sensor fusion, Field-Programmable Gate Arrays (FPGAs) are increasingly advantageous over traditional CPUs or GPUs due to their parallel processing capabilities and reconfigurability, allowing for custom, low-latency hardware acceleration for critical navigation and safety functions. Beyond remote oversight, a ground crew of "Fleet Attendants" and "Robot Delivery Specialists" performs essential hands-on work. In operational hubs, these teams are responsible for charging the robots, cleaning sensors, performing routine maintenance, and physically deploying the machines to their service areas at the start of a shift. This human-in-the-loop model is a practical solution to the "last-mile problem," which can account for over 50% of total shipping costs due to inefficiencies in the final leg of a delivery. While removing the driver for the bulk of the journey saves costs, the need for both remote and physical human support demonstrates the persistent complexity of navigating public urban environments. Interestingly, the human element extends beyond paid operators. A 2024 study revealed the public's "invisible work" in accommodating these robots. Researchers observed pedestrians subtly adjusting their paths, yielding right-of-way, and even signaling to the robots, unconsciously helping them navigate shared spaces and ensuring smoother operations.
