Robotics and computer vision / 2023
Brick-and-mortar retail still depends on manual stocktakes, paper customer requests, and inventory data that drifts out of sync with what is actually on the shelves. A robotic system that can see, navigate, and update business records continuously is operationally valuable but rarely shown end-to-end.
Built on the Husarion ROSbot 2 Pro platform in Gazebo, ARIR uses YOLO-family detectors with DeepSORT tracking to maintain a stable inventory view across frames, publishes product locations through ROS topics, synchronises state to an Odoo ERP via XML-RPC, and accepts customer requests that drive guided navigation through move_base.
My part
Decisions
Frame-by-frame detections produce visible double counts when products briefly leave and re-enter the field of view. DeepSORT track IDs plus a debounce buffer mean an item is only added or removed from inventory after stable observation, not on every frame.
Odoo state is the source of truth. The robot pushes diffs scoped to a stable product key so re-running a route or replaying a buffer cannot corrupt counts.
The navigation graph (gmapping + move_base) does not depend on the perception pipeline. Customer requests are independent goal sources, so perception failures degrade gracefully into guided navigation only.
Screens
SYSTEM ARCHITECTURE
RGB, depth, LiDAR and odometry
YOLO detection, DeepSORT and debouncing
gmapping, AMCL and TF transforms
Shelf proximity and map-relative state
Persistence and Odoo XML-RPC
move_base and customer product guidance
Gazebo simulation and shelf layout
Built with
Links
More work
Robot Guide DogReal-time ROS perception and audio-guidance system developed for a Unitree Go1 guide-dog concept, combining object detection, scene context, and synchronised spoken cues.Research prototypeVision Operations PlatformBrowser-to-inference platform for streaming camera frames to a FastAPI detection service, visualising detections in the browser, and monitoring streaming behaviour in real time.MaintainedPath Planning SimulatorInteractive browser-based simulator comparing A*, Dijkstra, and an RRT-style planner on a configurable grid with obstacle painting, start/goal placement, and live stats.Maintained