Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation
We present a novel method of measurement covariance estimation that models measurement uncertainty as a function of the measurement itself. Existing work in predictive sensor modeling outperforms conventional fixed models, but requires domain knowledge of the sensors that heavily influences the accu...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Published: |
IEEE
2020
|
| Online Access: | https://hdl.handle.net/1721.1/123673 |
| _version_ | 1826211615911968768 |
|---|---|
| author | Liu, Katherine Ok, Kyel Vega-Brown, William R Roy, Nicholas |
| author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
| author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Liu, Katherine Ok, Kyel Vega-Brown, William R Roy, Nicholas |
| author_sort | Liu, Katherine |
| collection | MIT |
| description | We present a novel method of measurement covariance estimation that models measurement uncertainty as a function of the measurement itself. Existing work in predictive sensor modeling outperforms conventional fixed models, but requires domain knowledge of the sensors that heavily influences the accuracy and the computational cost of the models. In this work, we introduce Deep Inference for Covariance Estimation (DICE), which utilizes a deep neural network to predict the covariance of a sensor measurement from raw sensor data. We show that given pairs of raw sensor measurement and ground-truth measurement error, we can learn a representation of the measurement model via supervised regression on the prediction performance of the model, eliminating the need for hand-coded features and parametric forms. Our approach is sensor-agnostic, and we demonstrate improved covariance prediction on both simulated and real data. Keywords: robot sensing systems; measurement uncertainty; measurement errors; covariance matrices; predictive models; estimation; neural networks |
| first_indexed | 2024-09-23T15:08:48Z |
| format | Article |
| id | mit-1721.1/123673 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T15:08:48Z |
| publishDate | 2020 |
| publisher | IEEE |
| record_format | dspace |
| spelling | mit-1721.1/1236732022-10-02T00:54:56Z Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation Liu, Katherine Ok, Kyel Vega-Brown, William R Roy, Nicholas Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory We present a novel method of measurement covariance estimation that models measurement uncertainty as a function of the measurement itself. Existing work in predictive sensor modeling outperforms conventional fixed models, but requires domain knowledge of the sensors that heavily influences the accuracy and the computational cost of the models. In this work, we introduce Deep Inference for Covariance Estimation (DICE), which utilizes a deep neural network to predict the covariance of a sensor measurement from raw sensor data. We show that given pairs of raw sensor measurement and ground-truth measurement error, we can learn a representation of the measurement model via supervised regression on the prediction performance of the model, eliminating the need for hand-coded features and parametric forms. Our approach is sensor-agnostic, and we demonstrate improved covariance prediction on both simulated and real data. Keywords: robot sensing systems; measurement uncertainty; measurement errors; covariance matrices; predictive models; estimation; neural networks United States. National Aeronautics and Space Administration (Award NNX15AQ50A) United States. Defense Advanced Research Projects Agency (Contract HR0011-15-C-0110) 2020-01-24T15:21:14Z 2020-01-24T15:21:14Z 2018-09-13 2018-05 Article http://purl.org/eprint/type/ConferencePaper 9781538630815 9781538630808 9781538630822 2577-087X https://hdl.handle.net/1721.1/123673 Liu, Katherine et al. "Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation." 2018 IEEE International Conference on Robotics and Automation (ICRA), May 21-25, 2018, Brisbane, Queensland, Australia, IEEE, 2018 http://dx.doi.org/10.1109/icra.2018.8461047 2018 IEEE International Conference on Robotics and Automation (ICRA) Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf IEEE Katherine Liu |
| spellingShingle | Liu, Katherine Ok, Kyel Vega-Brown, William R Roy, Nicholas Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation |
| title | Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation |
| title_full | Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation |
| title_fullStr | Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation |
| title_full_unstemmed | Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation |
| title_short | Deep Inference for Covariance Estimation: Learning Gaussian Noise Models for State Estimation |
| title_sort | deep inference for covariance estimation learning gaussian noise models for state estimation |
| url | https://hdl.handle.net/1721.1/123673 |
| work_keys_str_mv | AT liukatherine deepinferenceforcovarianceestimationlearninggaussiannoisemodelsforstateestimation AT okkyel deepinferenceforcovarianceestimationlearninggaussiannoisemodelsforstateestimation AT vegabrownwilliamr deepinferenceforcovarianceestimationlearninggaussiannoisemodelsforstateestimation AT roynicholas deepinferenceforcovarianceestimationlearninggaussiannoisemodelsforstateestimation |