WiFi-communicated Cooperative Platoon Running of Multiple Mobile Robots Using Image Machine Learning with FOMO*

In this paper, we propose a method to realize cooperative platoon running by multiple robots through image processing using machine learning. This study uses the learning model FOMO(Faster objects, More objects) provided by Edge Impulse as a method for object detection. FOMO simplifies object detection by aiming to find the geometric center position of objects, thereby reducing memory usage. This allows for faster image processing by the microcomputer. By applying the leading-edge digital technology, we propose a distance estimation method between two small mobile robots while cooperative running, in which real-time computing is conducted by using image processing by machine learning and detection of two characters attached to a preceding robot. Using this method, an estimation model was created to infer the distance between robots from the information of object detection. In addition, all microcomputers used in the robots are equipped with ESP32. This made it possible to use ultra-fast wireless communication (ESP-NOW) with multiple units and send the necessary information to the designated robot. This communication method does not require a router or access point, and it is possible to connect multiple slave devices to one master device. The combination of these image processing and wireless communications enables straight-line tracking by multiple robots. In addition, by ArUco markers on the robots, running path in experiments could be obtained and tracking performance could be evaluated. Finally, we demonstrate a platoon running method and easily applicable control law for small robots using FOMO, and verify that it is easy to estimate the distance and position of the preceding robot.