Direct orientation estimation through inertial odometry based on a deep transformer model
0
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
Studies on the inertial measurement unit (IMU), a fundamental localization solution for mobile devices in inertial odometry have been conducted. Most inertial odometry approaches focus on the two-dimensional space; however, technological advancements demand the accurate measurement of movements in the three-dimensional (3D) space, making 3D inertial odometry algorithms essen-tial. Three-dimensional inertial odometry calculates the relative pose based on IMU-measured data, and the estimated pose denotes the displacement encountered by the sensor within a unit of time. Subsequently, the trajectory is generated through integration. However, many existing approaches are characterized by drift errors owing to the integration processes utilized to estimate position and orienta-tion. We propose an extended direct orientation inertial odometry network (DO IONet) that directly estimates the orientation to over-come drift errors, improving the estimation performance of the 3D translation vector. The proposed approach calculates the orientation by inputting linear acceleration, gyroscope, gravity acceleration, and geomagnetic values, overcoming drift errors associated with ori-entation estimation. The extended DO IONet comprises an encoder for local feature extraction and a decoder for sequential feature extraction. The proposed model doesn't require structural initialization and doesn't cause drift error because an integration process is not required to estimate the orientation.
Abstract
Studies on the inertial measurement unit (IMU), a fundamental localization solution for mobile devices in inertial odometry have been conducted. Most inertial odometry approaches focus on the two-dimensional space; however, technological advancements demand the accurate measurement of movements in the three-dimensional (3D) space, making 3D inertial odometry algorithms essen-tial. Three-dimensional inertial odometry calculates the relative pose based on IMU-measured data, and the estimated pose denotes the displacement encountered by the sensor within a unit of time. Subsequently, the trajectory is generated through integration. However, many existing approaches are characterized by drift errors owing to the integration processes utilized to estimate position and orienta-tion. We propose an extended direct orientation inertial odometry network (DO IONet) that directly estimates the orientation to over-come drift errors, improving the estimation performance of the 3D translation vector. The proposed approach calculates the orientation by inputting linear acceleration, gyroscope, gravity acceleration, and geomagnetic values, overcoming drift errors associated with ori-entation estimation. The extended DO IONet comprises an encoder for local feature extraction and a decoder for sequential feature extraction. The proposed model doesn't require structural initialization and doesn't cause drift error because an integration process is not required to estimate the orientation.
참고문헌 (34)
- J. Wendel, "An integrated GPS", MEMS-IMU navigation system for an au-tonomous helicopter, 2006, pp.10(6), 527~533
- J. H. Seong, "Selective unsupervised learn-ing-based Wi-Fi fingerprint system using autoencoder and GAN", IEEE Internet of Things Journal, 7(3), pp.1898~1909 (2020)
- J. H. Seong, "Environment adaptive localiza-tion method using Wi-Fi and Bluetooth low energy", Wire-less Personal Communications, 99, pp.765~778 (2018)
- S. A. Mohamed, "A survey on odometry for autonomous navigation systems", IEEE ac-cess, 7, pp.97466~97486 (2019)
- N. Yang, "D3vo: Deep depth, deep pose and deep uncertainty for mo-nocular visual odometry", Proceedings of the IEEE, 2020, 1278~1289
- A. A. Deraz, "Deep learning based on LSTM model for enhanced visual odometry navigation sys-tem", Ain Shams Engineering Journal, 14(8), pp.102050~ (2023)
- Y. Zhou, "Event-based stereo vis-ual odometry", IEEE Transactions on Robotics, 37(5), pp.1433~1450 (2021)
- R. Duan, "Stereo orientation prior for UAV robust and accurate visual odometry", IEEE, 2022, pp.27(5), 3440~3450
- J. Zhang, "Low-drift and real-time lidar odom-etry and mapping", Autonomous Robots, 41, pp.401~416 (2017)
- Q. Li, "Lo-net: Deep real-time lidar odometry", Proceedings of the IEEE, 2019, 8465~8474
- P. Dellenbach, "CT-ICP: Real-time elastic LiDAR odometry with loop clo-sure", 2022 International Conference on Robotics and Automation (ICRA), pp.5580~5586 (2022)
- Y. S. Park, "PhaRaO: Direct radar odometry using phase correlation", 2020 IEEE Interna-tional Conference on Robotics and Automation (ICRA), pp.2617~2623 (2020)
- D. Adolfsson, "CFEAR radarodometry - Con-servative filtering for efficient and accurate radar odometry", 2021 IEEE, 2021, 5462~5469
- P. C. Kung, "A normal distri-bution transform-based radar odometry designed for scan-ning and automotive radars", 2021 IEEE International Conference on Robotics and Automation (ICRA), pp.14417~14423 (2021)
- S. Y. Cho, "A calibration technique for a two-axis magnetic compass in telematics devices", ETRI Journal, 27(3), pp.280~288 (2005)
- A. A. A. Rahni, "Obtaining translation from a 6-DOF MEMS IMU—An overview", 2007 Asia-Pacific Conference on Applied Electromagnetics, pp.1~5 (2007)
- X. Chen, "Novel hybrid of strong tracking Kalman filter and wavelet neural network for GPS", INS during GPS outages, 2013, pp.46(10), 3847~3854
- S. Y. Cho, "Robust positioning technique in low-cost DR", GPS for land navigation, 2006, pp.55(4), 1132~1142
- P. G. Savage, "Strapdown inertial navigation integration al-gorithm design part 1: Attitude algorithms", Journal of Guidance, Control, and Dynamics, 21(1), pp.19~28 (1998)
- P. G. Savage, "Strapdown inertial navigation integration al-gorithm design part 2: Velocity and position algorithms", Journal of Guidance, Control, and dynamics, 21(2), pp.208~221 (1998)
- W. Gao, "Adaptive Kalman filter-ing with recursive noise estimator for integrated SINS", DVL systems, 2015, pp.68(1), 142~161
- Z. Zheng, "An eight-position self-calibration method for a dual-axis rotational inertial naviga-tion system", Sensors and Actuators A: Physical, 232, pp.39~48 (2015)
- W. Kang, "SmartPDR: Smartphone-based pe-destrian dead reckoning for indoor localization", IEEE Sen-sors Journal, 15(5), pp.2906~2916 (2015)
- X. Hou, "Pedestrian dead reckoning with wearable sensors: A systematic review", IEEE Sensors Journal, 21(1), pp.143~152 (2021)
- X. Chen, "A novel calibration method for tri-axial magnetometers based on an expanded error model and a two-step total least square algorithm", Mobile Networks and Applications, 27(2), pp.794~805 (2022)
- H. Yan, "RIDI: Robust IMU double integration", Proceedings of the European Con-ference on Computer Vision (ECCV), pp.641~656 (2018)
- C. Chen, "Deep neural network based inertial odometry us-ing low-cost inertial measurement units", IEEE Transac-tions on Mobile Computing, 20(4), pp.1351~1364 (2021)
- C. Chen, "Deep-learning-based pedestrian inertial naviga-tion: Methods, data set, and on-device inference", IEEE In-ternet of Things Journal, 7(5), pp.4431~4441 (2020)
- M. Zhang, "IMU data pro-cessing for inertial aided navigation: A recurrent neural net-work based approach", 2021 IEEE International Confer-ence on Robotics and Automation (ICRA), pp.3992~3998 (2021)
- J. P. Silva do Monte Lima, "End-to-end learning framework for IMU-based 6-DOF odometry", Sensors, 19(17), pp.3777~ (2019)
- W. Y. Kim, "Nine-Axis IMU-based extended inertial odometry neural network", Expert Systems with Applications, 178, pp.115075~ (2021)
- H. I. Seo, "DO IONet: 9-axis IMU-based 6-DOF odometry framework using a neu-ral network for direct orientation estimation", IEEE Access, 11, pp.55380~55388 (2023)
- A. Kendall, "Multi-task learning us-ing uncertainty to weigh losses for scene geometry and se-mantics", Proceedings of the IEEE Conference on Com-puter Vision and Pattern Recognition, pp.7482~7491 (2018)