Leveraging Self-Attention Mechanism for Attitude Estimation in Smartphones

Abstract

Inertial attitude estimation is a crucial component of many modern systems and applications. Attitude estimation from commercial-grade inertial sensors has been the subject of an abundance of research in recent years due to the proliferation of Inertial Measurement Units (IMUs) in mobile devices, such as the smartphone. Traditional methodologies involve probabilistic, iterative-state estimation; however, these approaches do not generalise well over changing motion dynamics and environmental conditions, as they require context-specific parameter tuning. In this work, we explore novel methods for attitude estimation from low-cost inertial sensors using a self-attention-based neural network, the Attformer. This paper proposes to part ways from the traditional cycle of continuous integration algorithms, and formulate it as an optimisation problem. This approach separates itself by leveraging attention operations to learn the complex patterns and dynamics associated with inertial data, allowing for the linear complexity in the dimension of the feature vector to account for these patterns. Additionally, we look at combining traditional state-of-the-art approaches with our self-attention method. These models were evaluated on entirely unseen sequences, over a range of different activities, users and devices, and compared with a recent alternate deep learning approach, the unscented Kalman filter and the iOS CoreMotion API. The inbuilt iOS had a mean angular distance from the true attitude of 117.31∘, the GRU 21.90∘, the UKF 16.38∘, the Attformer 16.28∘ and, finally, the UKF–Attformer had mean angular distance of 10.86∘. We show that this plug-and-play solution outperforms previous approaches and generalises well across different users, devices and activities.

RIOT: Recursive Inertial Odometry Transformer for Localisation from Low-Cost IMU Measurements

“…The ARIOT model is a hierarchical Transformer; it differs from RIOT by the incorporation of an additional, internal attitude estimation network that regresses the orientation of the IMU from the sensor measurements. This subsystem benefits from the self-attention-based framework design in [ 34 ]. However, we were able to formulate a new loss function, which, to the best of our knowledge, is absent in the literature.…”

“…The Attitude Recursive Inertial Odometry Transformer is a hierarchical framework composed of two self-attention-based encoder-decoder networks, depicted in Figure 1 and Figure 2 . The foundation for the initial network is based on previous work [ 34 ] and functions to regress attitude estimation from 9D inertial measurements (from Equations ( 1 )–( 3 )). This allows for the componential estimation of both attitude and position estimation in a single framework, providing a robust solution for inertial odometry.…”

Generalizable End-to-End Deep Learning Frameworks for Real-Time Attitude Estimation Using 6dof Inertial Measurement Units}

“…In Table 1, we summarized some related works in the navigation field using deep learning. [49] 2015/12 Vision Relocalization VINet [50] 2017/02 Vision + Inertial Visual Inertial Odometry DeepVO [51] 2017/05 Vision Visual Odometry VidLoc [52] 2017/07 Vision Relocalization IONet [40] 2018/02 Inertial Only Inertial Odometry UnDeepVO [53] 2018/05 Vision Visual Odometry VLocNet [54] 2018/05 Vision Relocalization, Odometry RIDI [55] 2018/09 Inertial Only Inertial Odometry SIDA [56] 2019/01 Inertial Only Domain Adaptation VIOLearner [57] 2019/04 Vision + Inertial Visual Inertial Odometry RINS-W [58] 2019/05 Inertial Only Inertial Odometry SelectFusion [59] 2019/06 Vision + Inertial + LIDAR VIO and sensor Fusion LO-Net [60] 2019/06 LIDAR LIDAR Odometry L3-Net [61] 2019/06 LIDAR LIDAR Odometry Lima et al [62] 2019/8 Inertial Inertial Odometry DeepVIO [63] 2019/11 Vision+Inertial Visual Inertial Odometry OriNet [35] 2020/4 Inertial Inertial Odometry Sorg [64] 2020/4 Inertial Pose Estimation GALNet [65] 2020/5 Inertial, Dynamic and Kinematic Autonomous Cars PDRNet [66] 2021/3 Inertial Pedestrian Dead Reckoning Kim et al [67] 2021/4 Inertial Inertial Odometry RIANN [34] 2021/5 Inertial Attitude Estimation CTIN [68] 2022/6 Inertial Inertial Odometry Xia et al [69] 2022/8 Inertial Human Pose Estimation Brotchie et al [70] 2022/11 Inertial Attitude Estimation…”

Recent Advancements in Deep Learning Applications and Methods for Autonomous Navigation: A Comprehensive Review

“…Another study presented an LSTM-based model, but was restricted to a single sampling rate and did not mention publicly available inertial datasets Narkhede et al (2021). A third study utilized a self-attention mechanism, but its model was trained and tested solely on the OxIOD dataset, limiting its applicability to specific motion types and sampling rates Brotchie et al (2022). Asgharpoor et al introduced three end-to-end learning frameworks that aimed to generalize across different environments and sensor sampling rates, although the use of IMU data windows in these frameworks may introduce delays Asgharpoor Golroudbari and Sabour (2023).…”