Improving EKF-Based IMU/GNSS Fusion Using Machine Learning for IMU Denoising
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
In the realm of navigation systems, Inertial Measurement Unit (IMU) sensors play a pivotal role. The advent of Micro-Electro-Mechanical System (MEMS) sensors has introduced a lightweight and cost-effective alternative for IMUs. However, MEMS IMUs come with the challenge of larger stochastic errors that accumulate over time, resulting in navigation drifts. To address this issue, the conventional approach involves fusing IMU with the Global Navigation Satellite System (GNSS) for reliable navigation. Nevertheless, this fusion setup fails in providing ubiquitous navigation during GNSS outage scenarios due to persistent IMU errors. In this paper, an efficient methodology is developed to mitigate navigation drifts by eliminating IMU errors using Light Gradient Boosting Machine (LightGBM) and Categorical Boosting (CatBoost) Machine Learning (ML) algorithms. In contrast to existing methodologies that employ high-end and expensive IMUs for training models to denoise low-cost MEMS IMUs, this paper proposes utilizing Inverse Kinematics (IK). This approach helps to derive clean IMU training data from the Position, Velocity, Attitude (PVA) values estimated through the Extended Kalman Filter (EKF) when GNSS is available and reliable. The distinctive advantage of the IK approach lies in its capacity to obtain real-time pseudo error-free IMU data without the necessity for high-end IMUs to train ML models. The proposed method undergoes testing in both Loosely coupled and Tightly coupled EKF scenarios using simulation and real dataset under varying GNSS outage durations. Comparisons are made between the denoised IMU signals and signal processing techniques such as Moving Average (MA) and Savitzky Golay (SG). Additionally, we present a comparative analysis of the proposed algorithms against Convolutional Neural Networks (CNN). Results demonstrate a noteworthy enhancement in position, velocity, and orientation estimation. Furthermore, the computation time required for model training and prediction across various algorithms is analyzed. The outcomes prove the superiority of the proposed tree-based algorithms over the conventional filtering methods and CNN in denoising IMU and improving the navigation results.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it