The customer needed a way to inspect sites while tracking equipment location in environments where GPS is unavailable, such as tunnels and indoor facilities. However, GPS-dependent operations could not provide position estimation in these settings, making it difficult to link measurement data with actual movement on site.

When multiple tools are required to check and compare point-cloud data, the validation process becomes complicated, and the burden of continuously evaluating data collected on site increases. The customer needed a configuration that could handle position estimation, map generation, and review as one continuous workflow rather than separate tasks.

The goal of this project was to enable relative positioning and 3D mapping of on-site conditions in environments where GPS is unavailable by using LiDAR together with an IMU and camera.

Another priority was to make the collected data easier to use in later processes, creating conditions that would allow point clouds to be reviewed and validated more efficiently. The goal was not simply to connect sensors, but to build a foundation that could support an integrated workflow from on-site measurement through to verification.

SMILE was responsible for the full scope of development, from overall system architecture design to the integration of LiDAR, IMU, and cameras; building the ROS2-based processing platform; point-cloud processing; SLAM integration; map data storage; visualization through a Web viewer; and testing and tuning.

Because this was a highly R&D-oriented project, our work went beyond implementation to include designing an architecture that would be easy to validate and improve iteratively. Rather than optimizing sensor input, position estimation, and visualization in isolation, we developed the overall system configuration step by step so that it would function effectively as a whole.

In this project, given the constraint that GPS could not be used, we adopted a configuration that improves the stability of position estimation by combining LiDAR with an IMU and camera rather than relying on LiDAR alone. By integrating sensor data and using SLAM to handle self-positioning and the surrounding environment simultaneously, the system is better suited to mapping while in motion rather than one-off measurements.

Another important point is that the development was not limited to the algorithm itself; it was designed with ease of review in mind. By using RViz2, an internal review interface, and a Web viewer, SMILE organized processing results in a form that multiple stakeholders could easily understand. This created an environment that supports smoother alignment between technical validation and the envisioned operational use.

The main challenge was to keep SLAM running reliably in an environment where GPS was unavailable, while properly synchronizing LiDAR, IMU, and camera data and layering point clouds with minimal distortion. Because each sensor captures data at different timings and with different characteristics, simply lining up the data can easily lead to unstable position estimation and inconsistent map quality.

SMILE combined FAST-LIO with related processing while refining the data-flow design and sensor integration on ROS2, repeatedly validating and tuning the system. It also established a structure that makes results easier to visualize, enabling improvements while checking how point clouds overlap and behave. This created an environment for continuously improving accuracy and stability while maintaining a foundation that supports efficient technical validation.

Through this initiative, we established a configuration that supports 3D map generation and relative position verification even in environments where GPS cannot be used. It also reduces reliance on multiple standalone tools for point cloud review, creating a foundation that makes it easier to evaluate mapping results and drive improvement cycles.

For this project, we prioritized a structure that allowed the Japan and Vietnam teams to divide responsibilities while advancing development as an R&D initiative built around validation. Requirements and validation points were aligned between the two sides, and the BrSE bridged specification understanding and communication, enabling implementation and verification to proceed while minimizing gaps in understanding.

In technical validation projects, it is important to share not only implementation results but also progress along the way. SMILE maintained a continuous cycle of progress reporting, sharing validation results, conducting reviews, and organizing test perspectives, enabling accuracy and stability to be improved step by step. This approach established a development structure that makes it easier to validate and improve solutions, even for highly uncertain themes.

3D mapping in environments where GPS is unavailable cannot be achieved effectively by treating sensor integration, position estimation, point cloud processing, and visualization as separate tasks. In this project, SMILE designed and implemented them as an integrated system, building a foundation that could be repeatedly validated and refined toward practical use.

For companies in construction, infrastructure, manufacturing, and other sectors that need to work with spatial information in environments where GPS cannot be relied on, this case offers a practical reference for planning how to apply LiDAR and SLAM.