DustCart: The On-Demand Waste Collection Robot

Overview

DustCart was one of two autonomous platforms developed within the EU FP6 DustBot project (2006–2009). Designed for on-demand household waste collection in pedestrian areas, it represented one of the earliest attempts to deploy an autonomous service robot in a real urban setting with real users.

Design and Dimensions

DustCart weighed approximately 60 kilograms and stood roughly 1.2 metres tall — about the height of a parking meter. Its footprint was compact enough to navigate streets as narrow as 2 metres, which was a hard requirement for operating in Peccioli‘s medieval town centre.

The robot’s chassis was designed by Robotnik Automation in Valencia, Spain, and integrated with navigation, communication, and waste-handling subsystems developed by the broader consortium. It moved on wheels using differential drive — two independently powered wheels plus casters for stability — giving it the ability to turn on the spot in tight spaces.

Navigation System

DustCart’s navigation relied on three primary sensor inputs:

• Differential GPS — a local base station provided real-time corrections, achieving ~2 cm positioning accuracy
• Laser rangefinder — a 2D LiDAR unit (likely SICK LMS series) for obstacle detection and map-based localisation
• Inertial measurement unit — accelerometers and gyroscopes for bridging GPS gaps in narrow streets

These inputs were fused using an Extended Kalman Filter to produce a continuous position estimate. The navigation architecture used a three-layer hierarchy: topological global planning, metric trajectory following, and reactive obstacle avoidance.

Operational Model

DustCart’s defining feature was its on-demand service model. Residents telephoned a dedicated number to request waste collection. The request was routed to a central control system, which dispatched the nearest available robot to the caller’s address. The robot navigated autonomously to the pickup location, accepted the waste, and transported it to a collection depot.

This model was structurally similar to what modern delivery robots do today — receive a request, navigate to a location, complete a task, return to base — but in reverse (collecting rather than delivering).

Communication

DustCart maintained connectivity through a hybrid communication architecture: a Wi-Fi mesh network with cellular fallback. The mesh provided primary data links within the operating area; cellular connectivity served as a backup in areas beyond mesh coverage.

Field Trials

DustCart was primarily tested in Peccioli, Italy. The trials demonstrated autonomous navigation in unstructured pedestrian environments, successful completion of on-demand collection tasks, and acceptable interaction with non-expert users. Limitations included restricted operating hours (the robot could not operate in darkness due to GPS requirements and safety considerations) and the need for a flat, unobstructed route between pickup and depot.

Legacy

DustCart did not become a commercial product, but it proved that on-demand autonomous services in pedestrian environments were technically feasible. Its DNA is visible in the Starship and Nuro platforms that followed a decade later.

References

• Ferri, G. et al. (2011). “DustCart, an autonomous robot for door-to-door garbage collection.” IEEE ICRA 2011.
• Salvini, P. et al. (2011). “The Robot DustCart.” IEEE Robotics & Automation Magazine.