Back to Insights

What Is Multi-SLAM Technology (Multi-Sensor Fusion SLAM)?

Tech Insights5 min read

What is Multi-SLAM Technology? The limitations of single sensor SLAM technology and advantages and effects of multi-SLAM.

Simultaneous Localization and Mapping (SLAM) technology refers to mobile robots using their own sensors to build environmental maps and use environmental information for autonomous positioning.

There are two mainstream technical solutions - visual SLAM and LiDAR SLAM. However, they have their own shortcomings when facing complex environments: visual SLAM systems are easily affected by changes in the visual environment, while LiDAR SLAM systems may experience accuracy degradation in environments without feature points. XGRIDS is the first company in the mobile SLAM scanning industry to apply multi-sensor fusion SLAM (i.e. Multi-Slam) technology to achieve more stable and higher quality data in complex scenes.

Limitations of Single Sensor SLAM Technology

  • Limitations of Visual SLAM: Visual SLAM systems are susceptible to changes in the visual environment, such as repeated textures, lighting changes, dynamic environments, etc. These factors may cause feature point matching failures, affecting the stability and accuracy of the system.Vision sensors are essentially angle measurement sensors that cannot directly measure distance information. They need to reconstruct the distance of features from multiple views, which increases computational complexity and error risk, resulting in poor measurement accuracy.
  • Limitations of Lidar SLAM: Lidar SLAM systems may experience accuracy degradation or even mapping failure in environments with a monotonous structure or unclear features, such as tunnels or highways.LiDAR data will be distorted when the sensor carrier is dynamic, resulting in a decrease in SLAM accuracy.
  • In order to solve the limitations of single sensors, we further explore the technical path of multi-sensor fusion SLAM.
Surveyor walking beneath a concrete overpass while capturing data with an XGRIDS handheld 3D scanner
Surveyor walking beneath a concrete overpass while capturing data with an XGRIDS handheld 3D scanner

Advantages and Effects of Multi-SLAM

There is good complementarity between LiDAR, cameras, and Inertial Measurement Units (IMUs). IMU measurement is not affected by environmental characteristics, and inertial navigation systems can estimate the full parameters of speed, position, and attitude based solely on the inertial information generated by carrier motion. The motion parameters estimated by inertial navigation can be used to correct the distortion of LiDAR data, compensate for the missing scale information of monocular vision, etc.; while the carrier motion measured by LiDAR SLAM and visual SLAM can correct the cumulative error of inertial navigation systems. Multi-sensor fusion SLAM technology improves the performance of SLAM systems by combining the complementarity of different sensors.

  1. Improve accuracy and robustness: Multi-sensor fusion can utilize complementary information from different sensors, such as motion information provided by IMU, which can correct the distortion of LiDAR SLAM data and compensate for the missing scale information of visual SLAM, thereby improving the accuracy and robustness of the system.
  2. Greater environmental adaptability: Multi-sensor SLAM systems can operate under a wider range of environmental conditions, such as in scenarios where visual environmental features are scarce or LiDAR data accuracy is declining. By fusing other sensor data, the system can still maintain good performance.
  3. Improve system reliability: In the event that some sensors fail or the data is unreliable, other sensors can provide backup information to ensure the continuous operation of the SLAM system.
  4. Enhance system dynamics: Multi-sensor fusion can render higher frequency updates and more accurate dynamic models, adapting to fast-moving carriers and systems.

Difficulties in Multi-Sensor SLAM Fusion

Although multi-sensor fusion SLAM technology has obvious advantages, there are also many difficulties. These are also the directions that XGRIDS continues to invest in research and development efforts to solve.

  • Real-time: While ensuring the accuracy of the algorithm, it is also necessary to consider the real-time performance of the algorithm to meet the needs of fast-moving platforms. In fact, the original meaning of SLAM - simultaneous localization and mapping - defines that SLAM must process results in real-time. As the first team provides truly mobile SLAM scanner in the industry, XGRIDS has already achieved real-time computing on board by the first generation product. Currently, the latest products have been optimized to achieve real-time data quality that are close to the post-processing effects of the previous generation products.
Colored point cloud comparison of European facades showing real-time Lixel L2 Pro output versus previous-generation results
Colored point cloud comparison of European facades showing real-time Lixel L2 Pro output versus previous-generation results

Post-processing data of previous product

  • Sensor calibration: It is necessary to accurately calibrate the relative position and time synchronization between different sensors, which is a prerequisite for effective fusion of sensor data. The calibration process is complex and requires high-precision calibration equipment or algorithms.
  • Data fusion algorithm: The fusion of multi-sensor data requires complex algorithms to handle the data characteristics of different sensors, such as the data frequency, timestamp alignment, and data format of different sensors.
  • Consistency of information: Ensuring the consistency of information from different sensors during the fusion process is a challenge, especially when there are differences in sensor observations.
  • Computing resource requirements: Multi-SLAM systems usually require more computing resources, including processor performance and memory. XGRIDS has long experience been in the field of distributed high-performance computing, using EdgeComputing to improve the data processing capability on board and achieve real-time reconstruction of fusing multiple sensors.
  • Algorithm complexity: Designing an effective multi-SLAM algorithm requires considering multiple factors, such as the dynamic characteristics of sensors and environmental changes, which increases the complexity of algorithm design.

In the field of mobile SLAM scanning, XGRIDS has led the innovation of multi-SLAM technology through unremitting technical pursuit and practice. Faced with the challenges of visual SLAM and LiDAR SLAM in complex environments, XGRIDS not only proposed a multi-sensor solution that integrates LiDAR, camera, and IMU in concept, but also achieved technological iteration from the first generation product Lixel L1 to the latest Lixel L2 Pro in practice, significantly improving the accuracy, robustness, and environmental adaptability of the system. By solving key difficulties such as real-time performance, sensor calibration, data fusion algorithms, information consistency, and computing resource requirements, XGRIDS demonstrates its pursuit of technological breakthroughs and down-to-earth approach, continuously promoting the development of 3D handheld SLAM scanning equipment towards more efficient and intelligent directions, and setting a new technological benchmark for the industry.

Share this article

More Articles