Abstract:Abstract Driver steering intention prediction provides an augmented solution to the design of an onboard collaboration mechanism between human driver and intelligent vehicle. In this study, a multi-task sequential learning framework is developed to predict future steering torques and steering postures based on upper limb neuromuscular electromyography signals. The joint representation learning for driving postures and steering intention provides an in-depth understanding and accurate modelling of driving steering behaviours. Regarding different testing scenarios, two driving modes, namely, both-hand and single-right-hand modes, are studied. For each driving mode, three different driving postures are further evaluated. Next, a multi-task time-series transformer network (MTS-Trans) is developed to predict the future steering torques and driving postures based on the multi-variate sequential input and the self-attention mechanism. To evaluate the multi-task learning performance and information-sharing characteristics within the network, four distinct two-branch network architectures are evaluated. Empirical validation is conducted through a driving simulator-based experiment, encompassing 21 participants. The proposed model achieves accurate prediction results on future steering torque prediction as well as driving posture recognition for both two-hand and single-hand driving modes. These findings hold significant promise for the advancement of driver steering assistance systems, fostering mutual comprehension and synergy between human drivers and intelligent vehicles.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict the driver's future steering torque and steering posture by combining neuromuscular dynamics and multi - task time - series Transformer (MTS - Trans), so as to improve the collaborative efficiency and safety of the human - machine co - driving system. Specifically, the research aims to: 1. **Develop a multi - task sequence learning framework**: This framework can predict future steering torque and steering posture based on the electromyogram signals of the upper - limb neuromuscular. By jointly representing and learning the driving posture and steering intention, it provides in - depth understanding and accurate modeling of the driving steering behavior. 2. **Study the influence of different driving modes and postures**: The paper explores the influence of different driving postures on the prediction of steering intention under the two - handed and one - handed driving modes. Specifically, for the two - handed driving mode, the driving postures at 3 o'clock, 10 - 10 o'clock and 12 o'clock are studied; for the one - handed driving mode, the driving postures at 3 o'clock, 130 o'clock and 12 o'clock are studied. 3. **Design and evaluate the multi - task time - series Transformer network**: A multi - task time - series Transformer network (MTS - Trans) is developed to predict future steering torque and driving posture based on multivariate sequence input and self - attention mechanism. In order to evaluate the multi - task learning performance and the information - sharing characteristics within the network, four different two - branch network architectures are evaluated. 4. **Experimental verification**: Through the experiment based on the driving simulator, 21 participants were tested to verify the accuracy of the proposed model in predicting future steering torque and identifying driving postures. These findings are of great significance for promoting the development of the driver's steering assist system and help to enhance the mutual understanding and cooperation between human drivers and intelligent vehicles. ### Key contributions 1. **Multi - task time - series model**: A multi - task time - series model is proposed to predict the driver's steering intention and identify the driving posture. This framework establishes the connection between the driver's neuromuscular dynamics and the future steering torque. This is the first attempt to jointly model the steering posture and steering torque prediction and further explore the design of the multi - task learning framework. 2. **Influence of different driving modes and postures**: Based on the proposed model, the influence of different driving modes and postures on the prediction of steering intention is studied, revealing the unique steering patterns under different driving postures. 3. **Quantitative analysis and comparison**: Quantitative analysis and comparison methods are proposed for different driving modes, postures and features to evaluate the performance of the MTS - Trans framework. ### Method overview 1. **Experimental design**: A six - degree - of - freedom driving simulator is used for the experiment, and the electromyogram (EMG) signals and steering torque data of the participants are collected. The experiment includes two driving modes (two - handed and one - handed), and each mode has three different driving postures. 2. **Signal processing**: The EMG signals are pre - processed, including low - pass filtering and normalization. Sequence data are generated using a sliding window, and each sequence has a length of 200 milliseconds, corresponding to 200 sampling points. 3. **Model architecture**: Four different MTS - Trans architectures are designed, and each architecture includes an input embedding layer, a Transformer encoder and a fully - connected prediction head. The model is optimized through a multi - task learning framework and trained by combining cross - entropy loss and mean - squared - error loss. ### Conclusion This research successfully predicts the driver's steering torque and driving posture by combining neuromuscular dynamics and multi - task time - series Transformer. This achievement not only helps to improve the performance of the driving assist system, but also provides new ideas for the human - machine interaction of future intelligent vehicles.

Driver Steering Behaviour Modelling Based on Neuromuscular Dynamics and Multi-Task Time-Series Transformer

Multi-task Driver Steering Behaviour Modeling Using Time-Series Transformer

Hybrid-Learning-Based Driver Steering Intention Prediction Using Neuromuscular Dynamics.

Continuous Driver Steering Intention Prediction Considering Neuromuscular Dynamics and Driving Postures

Pattern Recognition and Characterization of Upper Limb Neuromuscular Dynamics during Driver-Vehicle Interactions

Drivers' Manoeuvre Modelling and Prediction for Safe HRI

Data-driven Steering Torque Behaviour Modelling with Hidden Markov Models

Modeling Lateral Control Behavior of Driver for Manual Takeover of Automated Vehicles During Pedestrian Collision Avoidance

Personalized Driver Model and Its Application to Vehicle Testing

Model Predictive Control of Steering Torque in Shared Driving of Autonomous Vehicles.

Driver intention prediction based on multi-dimensional cross-modality information interaction

Modeling and analysis of driver behavior under shared control through weighted visual and haptic guidance

Deep Steering: Learning End-to-End Driving Model from Spatial and Temporal Visual Cues

A Unified Multi-scale and Multi-task Learning Framework for Driver Behaviors Reasoning

Modeling takeover time for non-driving related tasks using hybrid deep neural networks in conditionally automated vehicles

A self‐learning human‐machine cooperative control method based on driver intention recognition

Human-vehicle dynamic model with driver’s neuromuscular characteristic for shared control of autonomous vehicle

Driver lane change intention prediction based on topological graph constructed by driver behaviors and traffic context for human-machine co-driving system

Implementation of the Human-Like Lane Changing Driver Model Based on Bi-LSTM

Human-Like Decision Making of Artificial Drivers in Intelligent Transportation Systems: An End-to-End Driving Behavior Prediction Approach

Modeling Lateral Control Behaviors of Distracted Drivers for Haptic-Shared Steering System