EKF-Based Fusion of Wi-Fi/LiDAR/IMU for Indoor Localization and Navigation

Zeyi Li, Zhe Tang, Kyeong Soo Kim, Sihao Li, Jeremy S. Smith

TL;DR

This paper presents a novel indoor localization and navigation system that fuses Wi-Fi RSSI fingerprinting, LiDAR-based SLAM, and IMU data using an Extended Kalman Filter (EKF). The proposed EKF-based fusion significantly improves localization accuracy and robustness compared to individual sensor approaches, achieving mean 2D errors as low as 0.2449 m.

What Problem Does This Paper Address?

Accurate indoor localization is crucial for the growing deployment of robots, intelligent logistics, and mobile devices in smart buildings. Traditional GPS is unreliable indoors due to signal attenuation and multipath interference. Existing indoor localization methods, such as Wi-Fi RSSI fingerprinting, suffer from low accuracy and instability due to environmental changes. LiDAR-based SLAM offers high accuracy but is expensive and complex. IMUs provide high-frequency motion data but accumulate drift errors over time. This paper addresses the need for a robust, accurate, and cost-effective indoor localization system by fusing these complementary sensor modalities.

Methods

The core of the proposed system is an Extended Kalman Filter (EKF), a well-established algorithm for state estimation in non-linear systems. The EKF is used to fuse data from three sources:

  • DNN-Based Wi-Fi RSSI Fingerprinting: A deep neural network (DNN) is employed to process high-dimensional Wi-Fi RSSI vectors and estimate the 2D coordinates. This DNN acts as a coarse localization prior, refined by other sensors.
  • LiDAR-based SLAM (Gmapping): The Gmapping algorithm is used for simultaneous localization and mapping. It integrates 2D LiDAR scans with wheel odometry to build a consistent, high-resolution environment map and estimate the AGV's pose trajectory. This provides a more stable and accurate spatial representation than Wi-Fi alone.
  • IMU (Inertial Measurement Unit): The IMU provides high-frequency measurements of linear acceleration and angular velocity. This data is crucial for estimating the platform's orientation (attitude) and motion, compensating for the lower update rates of Wi-Fi and LiDAR and improving short-term localization continuity.

The EKF framework integrates these sensor inputs by:

  1. State Transition Model: Using IMU data to predict the AGV's next state (position, velocity, orientation).
  2. Observation Model: Incorporating the 2D location estimated by the DNN-based Wi-Fi fingerprinting as an observation.
  3. Update Step: Using the Kalman gain to fuse the predicted state with the observed location, correcting for errors and drift.

Datasets

The paper does not explicitly mention the creation or release of new datasets. Instead, it describes experiments conducted using the IR building at Xi'an Jiaotong-Liverpool University (XJTLU). The AGV platform used in the experiments collected data from Wi-Fi, LiDAR, and IMU sensors along predefined trajectories on the 6th, 7th, and 8th floors of this building.

Key Insights and Ideas

  • Synergistic Fusion: The paper demonstrates that fusing Wi-Fi, LiDAR, and IMU data through an EKF is significantly more effective than using any single sensor or a two-sensor combination. Each sensor compensates for the weaknesses of the others.
  • DNN as a Coarse Prior: Using a DNN for Wi-Fi fingerprinting provides a robust initial localization estimate, which is then refined by the more precise but potentially drift-prone LiDAR/IMU data.
  • Addressing Drift and Instability: The EKF's prediction-correction mechanism is key to suppressing the cumulative drift errors from IMUs and the instability and interference issues of Wi-Fi, leading to smoother and more accurate trajectories.
  • Path Invariance: The proposed method shows good directional symmetry and path invariance, meaning it performs consistently regardless of whether the AGV moves forward or backward along a path.

Key Results

The proposed EKF-based multi-sensor fusion framework achieved significantly lower mean 2D errors compared to individual methods:

  • Overall Mean 2D Errors: Ranged from 0.2449 m to 0.3781 m across different floors.
  • DNN-based Wi-Fi Fingerprinting: Mean 2D errors ranged from 0.5127 m to 1.3404 m, showing significant fluctuations, especially in areas with signal interference.
  • LiDAR/IMU Localization: Mean 2D errors ranged from 0.6233 m to 2.8803 m, indicating cumulative drift issues.
  • Specific Example (7th Floor): Wi-Fi error was 0.8792 m, LiDAR/IMU was 0.8553 m, while the EKF fusion achieved 0.2449 m.
  • Forward/Backward Paths: The EKF method maintained low errors (0.3672 m forward, 0.2851 m backward) while Wi-Fi and LiDAR/IMU showed higher and more variable errors.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that while the EKF effectively suppresses errors, it still faces challenges like data synchronization and coordinate system alignment, which are common in multi-sensor fusion.
  • The performance of Wi-Fi fingerprinting is still susceptible to severe signal fluctuations and multipath effects, though the fusion mitigates this.
  • LiDAR/IMU localization can still exhibit higher errors, particularly after sharp turns, due to inertial drift.
  • Future Work: Not explicitly detailed in the paper, but the success of this EKF-based fusion suggests potential for further exploration with more advanced deep learning architectures for sensor fusion, or integration with other sensor modalities for even greater robustness in highly complex or dynamic indoor environments.

Summary updated

Abstract

Conventional Wi-Fi received signal strength indicator (RSSI) fingerprinting cannot meet the growing demand for accurate indoor localization and navigation due to its lower accuracy, while solutions based on light detection and ranging (LiDAR) can provide better localization performance but is limited by their higher deployment cost and complexity. To address these issues, we propose a novel indoor localization and navigation framework integrating Wi-Fi RSSI fingerprinting, LiDAR-based simultaneous localization and mapping (SLAM), and inertial measurement unit (IMU) navigation based on an extended Kalman filter (EKF). Specifically, coarse localization by deep neural network (DNN)-based Wi-Fi RSSI fingerprinting is refined by IMU-based dynamic positioning using a Gmapping-based SLAM to generate an occupancy grid map and output high-frequency attitude estimates, which is followed by EKF prediction-update integrating sensor information while effectively suppressing Wi-Fi-induced noise and IMU drift errors. Multi-group real-world experiments conducted on the IR building at Xi'an Jiaotong-Liverpool University demonstrates that the proposed multi-sensor fusion framework suppresses the instability caused by individual approaches and thereby provides stable accuracy across all path configurations with mean two-dimensional (2D) errors ranging from 0.2449 m to 0.3781 m. In contrast, the mean 2D errors of Wi-Fi RSSI fingerprinting reach up to 1.3404 m in areas with severe signal interference, and those of LiDAR/IMU localization are between 0.6233 m and 2.8803 m due to cumulative drift.
BibTeX
@inproceedings{li2025ekf,
  author = {Zeyi Li and Zhe Tang and Kyeong Soo Kim and Sihao Li and Jeremy S. Smith},
  title = {{EKF-Based Fusion of Wi-Fi/LiDAR/IMU for Indoor Localization and Navigation}},
  booktitle = {CANDARW},
  year = {2025},
  url = {https://openreview.net/forum?id=I5290zln4t},
}
Figure