PERSOL CROSS TECHNOLOGY and Unseed Launch Joint R&D on Unmanned Construction Equipment Operations

prtimes.jp reported on Oct. 30, 2026, that PERSOL CROSS TECHNOLOGY and Unseed had launched joint research and development on unmanned construction equipment operations. We have not verified the primary source, so readers should consult the original for details.

The research focuses on automating not only the movement of dump trucks at construction sites but also the spreading of soil while they are moving. This presents a concrete test of approaches that handle travel and work separately. The companies say they will study the technology in simulation and are also considering tests with actual equipment, making it important to determine whether integrated control works on real machinery.

The research focuses on treating not only the travel paths of construction equipment but also the spreading of soil while moving as a single control task. The announcement says construction equipment often controls travel and work separately, and the companies are starting research to link those functions using VLA. However, the announcement marks the start of R&D; it does not confirm that unmanned operations have been achieved or that performance at a worksite has been demonstrated. The planned approach is to begin by building and validating a VLA model in simulation, combine that work with expertise in vehicle control, experiments and evaluation, and assess its potential application to actual equipment. A specific feature of the effort is that it includes vehicle-side controls and evaluation to connect the AI model to machine operation, rather than relying on the model alone. PERSOL CROSS TECHNOLOGY will contribute expertise in simulation, vehicle control, experiments and evaluation, while Unseed will contribute its AI model R&D capabilities. The roles described in the announcement span model development and the implementation and testing of machinery. Focusing on soil transport by dump truck and starting with the combined task of spreading soil while moving sets up a test of the ability to handle travel and work at the same time. By contrast, examples cited for the future, such as traveling to a designated location and unloading, and backfilling an area, were not described as completed initial research targets. Generating actions in response to natural-language instructions should be understood as a broader goal beyond the work now being undertaken. To assess the implications for the industry, the key questions are not only whether the technology works in simulation but also how it synchronizes travel and spreading on actual equipment and responds to changing conditions at a worksite. The announcement did not specify the model architecture, training data, evaluation metrics, timing for tests with actual equipment, or safety measures and conditions for human involvement at deployment. The next points to examine are therefore the differences between simulation and real-world results, the scope of the tasks covered and the criteria for achieving unmanned operations. The announcement also lists possible applications in automobiles, industrial machinery and robots, but does not establish that the technology will be applied to them.