Space Data and ACSL conducted verification of connecting autonomous control and image recognition functions verified on a digital twin to ACSL's domestically developed industrial drone, SOTEN (蒼天). Through integration between the digital twin and the actual drone, they aim to improve safety and reproducibility in real-world operations.
Using the digital twin environment of Sakai, Ibaraki Prefecture, Space Data and ACSL conducted verification of connecting drone control and recognition functions developed and tested in a virtual environment to SOTEN (蒼天). By combining digital twin technology with operational expertise, they established a process spanning verification of flight, recognition and control through evaluation using the actual drone.
On the digital twin, NVIDIA Isaac Sim was used to connect a vehicle model based on SOTEN to ground station software. Instructions such as takeoff, flight to designated points, turning and landing were reflected in the virtual drone, and a workflow was established for returning footage from its onboard camera to the operations interface. Functions including automatic hovering based on searches for color markers and people are currently under development, with the aim of verifying conditions consistently in the field and in the virtual environment.
For connection to the actual drone, they confirmed automatic control from takeoff through target detection and hovering, as well as the acquisition of gimbal and camera footage. The color-marker search developed in the virtual environment was applied to the actual drone, and efforts are continuing to identify and address differences between the actual drone and the simulation, including video latency, field of view and communications status. Going forward, they plan to increase verification that reproduces field conditions on an equivalent basis and expand into joint verification with companies and local governments.
Looking ahead, they aim to improve the accuracy of vehicle models and sensors by comparing actual-drone data with simulation results, and are considering expansion into fields including defense, disaster preparedness and social infrastructure. They will promote a co-creation approach that enables safer field verification while improving cost efficiency and reproducibility.