CCTV network steers 4 warehouse robots with no onboard navigation hardware

Researchers from the Oxford Robotics Institute (Luke Robinson, Benjamin Ramtoula, Anas Izaaryene, Paul Newman, Daniele De Martini) have published the first field demonstration of multi-robot planning and control using only a distributed CCTV camera network and off-board compute. The system drives multiple robots through a real warehouse using 30 security cameras spanning six 27-meter aisles, with robots carrying no task-specific navigation hardware. By operating entirely in image space over an uncalibrated, pixel-wise topological camera graph, the approach allows flexible camera placement and wide-area operation without traditional calibration.

The hierarchical planner first selects a sequence of cameras for each robot, then plans the robot's image-space motion through each camera's view. Coordination is achieved with a prioritized-then-joint strategy: overlapping camera regions are treated as shared resources held by one robot at a time, preventing collisions and deadlocks. In validation runs with four robots, the system reported mission times and coordination statistics, demonstrating that scalable multi-robot autonomy can be offloaded from onboard sensors to existing surveillance infrastructure and edge computing. This opens a practical path for logistics automation, retrofitting warehouses with minimal hardware changes.