Combain R&D

Combain invests substantially into positioning research to create practical and cost-efficient technologies that solve the need to track assets and people where GNSS is insufficient.

Project Background

Creating indoor navigation systems has historically been a labor-intensive process. It typically requires extensive manual work to develop accurate maps and define routes within buildings. This often involves detailed surveys, manual data collection, and significant time investments to ensure that navigation aids are precise and user-friendly. The Global Indoor Navigation (GIN) project was launched to address these challenges and transform indoor navigation through self-learning algorithms and machine-learning techniques.

Project Overview

The GIN project is an industrial positioning research initiative to develop a self-learning indoor navigation system that automatically characterizes building interiors using data from standard mobile devices. The primary objectives include:

• Automatic Building Characterization:
  Identifying the number of floors and locations of stairs, elevators, and escalators and creating simplified indoor maps.
• Building and Room Type Identification:
  Classifying buildings (e.g., residential, office, retail) and determining room types to enhance navigation relevance.

By achieving these goals, the project envisions enabling a multitude of valuable services, such as:

• Indoor Navigation:
  Providing users with accurate directions within complex indoor environments.
• Security Solutions:
  Enhancing safety protocols by understanding building layouts.
• Asset and Personnel Tracking:
  Locating individuals in need and tracking valuable assets efficiently.
• Building Change Detection:
  Monitoring modifications within structures to maintain up-to-date navigation data.
• Bottleneck Identification:
  Recognizing areas prone to congestion to improve accessibility and flow.

Technological Foundation

The GIN project builds upon advancements from a prior positioning research endeavor titled “Global Indoor Positioning in 3D.” This earlier project leveraged Simultaneous Localization and Mapping (SLAM) technology to automatically establish accurate indoor positioning using existing Wi-Fi access points and Bluetooth beacons. Integrating SLAM with the emerging Wi-Fi standard 802.11mc, which offers meter-level precision in indoor settings, presents new opportunities for innovative indoor navigation solutions.

Project Phases and Achievements

The GIN project has been structured into multiple phases, each focusing on specific aspects of development:

1. Phase 1 – Preparations and Initial Development: This phase involved laying the groundwork for the project, including setting objectives, assembling the research team, and initiating preliminary development tasks.
2. Phase 2 – Generating Initial Paths and Advancing Towards 1m Accuracy: Efforts were concentrated on developing algorithms capable of generating initial navigation paths within buildings and enhancing positioning accuracy to approximately one meter.
3. Phase 3—Automatic Route Extraction: A significant milestone was the development of methods to extract common routes within buildings automatically. By collecting extensive tracking data in various environments, such as a 400-square-meter office area equipped with multiple Wi-Fi access points and Bluetooth beacons, the team devised techniques to combine individual tracks into main paths and identify key nodes. This automatic extraction is crucial for enabling global indoor navigation without manual intervention.
4. Phase 4 – Multi-Floor Buildings and Artificial Neural Network Positioning: The focus shifted to addressing the complexities of multi-floor structures. By employing artificial neural networks, the project achieved a median positioning error of approximately 2-5 meters across various building types, including offices, residential areas, universities, and shopping malls. This represents a two to tenfold improvement over traditional trilateration methods.
5. Phase 5 – Integration and Demonstration: In the final phase, the project culminated in developing a comprehensive prototype for self-learning indoor navigation. Key components include:
   • Android Application with Indoor SDK: A new app facilitating easy data collection and survey building, enabling accurate indoor positioning within an hour.
   • Deep Neural Network Positioning Method: An advanced positioning technique providing superior accuracy without additional infrastructure.
   • Automatic Route Extraction Method: Algorithms that calculate entrances and common paths within buildings, achieving path accuracy between 2 and 10 meters.
   • Visualization Portal: A platform displaying building features, calculated paths, and demonstrating indoor navigation capabilities.

Future Prospects

The success of the GIN project has laid a robust foundation for future developments in indoor navigation. Based on GIN’s outcomes, the project team plans to continue positioning research and product development. Notably, the first indoor navigation solution will be integrated into the Traxmate IoT tracking platform, with deployment to a major customer anticipated before the end of 2022. Subsequent releases are expected to incorporate further research findings, enhancing the system’s capabilities.

Conclusion

The GIN project represents a significant leap forward in indoor navigation technology. Automating the characterization of building interiors and utilizing self-learning algorithms addresses the limitations of traditional manual methods. The project’s achievements demonstrate the feasibility of self-learning indoor navigation systems and pave the way for various applications that enhance indoor environments’ safety, efficiency, and user experience.