Abstract:There is evidence that the driving style of an autonomous vehicle is important to increase the acceptance and trust of the passengers. The driving situation has been found to have a significant influence on human driving behavior. However, current driving style models only partially incorporate driving environment information, limiting the alignment between an agent and the given situation. Therefore, we propose a situation-aware driving style model based on different visual feature encoders pretrained on fleet data, as well as driving behavior predictors, which are adapted to the driving style of a specific driver. Our experiments show that the proposed method outperforms static driving styles significantly and forms plausible situation clusters. Furthermore, we found that feature encoders pretrained on our dataset lead to more precise driving behavior modeling. In contrast, feature encoders pretrained supervised and unsupervised on different data sources lead to more specific situation clusters, which can be utilized to constrain and control the driving style adaptation for specific situations. Moreover, in a real-world setting, where driving style adaptation is happening iteratively, we found the MLP-based behavior predictors achieve good performance initially but suffer from catastrophic forgetting. In contrast, behavior predictors based on situationdependent statistics can learn iteratively from continuous data streams by design. Overall, our experiments show that important information for driving behavior prediction is contained within the visual feature encoder. The dataset is publicly available at

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the driving - style adaptability of autonomous vehicles in order to increase passengers' acceptance and trust in autonomous driving technology. Specifically, the author points out that current driving - style models largely ignore the influence of driving - environment information, which limits the matching degree between the agent (i.e., the autonomous driving system) and a given situation. Therefore, they propose a situation - aware driving - style model based on different visual - feature encoders (these encoders are pre - trained on fleet data) and predictors for specific drivers' driving behaviors. ### Main problems solved in the paper: 1. **Improve driving - style adaptability**: By considering driving - environment information, make the driving style of the autonomous driving system closer to the habits of human drivers, thereby enhancing passengers' comfort and trust in the autonomous driving system. 2. **Enhance the environmental adaptability of the model**: Existing driving - style models usually only contain static parameters and lack the adaptability to individual drivers. The method proposed in the paper can better adapt to different driving situations by combining visual - feature encoders and driving - behavior predictors. 3. **Improve driving - behavior prediction**: By using pre - trained visual - feature encoders and an unsupervised clustering method, the model can predict driving behaviors more accurately and form reasonable driving - situation clusters. ### Specific objectives: - **Propose a situation - aware driving - style adaptation method**: Utilize the representations learned from the driving environment to achieve dynamic adjustment of the driving style. - **Develop an interpretable clustering method**: Used to learn driving behaviors in specific situations and control the adaptation of the driving style. - **Provide a public data set**: Containing a large number of images and annotated driving - behavior data for research and verification. - **Introduce new evaluation metrics**: Such as the Entropy - based Cluster Specificity (ECS) measure, used to evaluate the specificity of the identified driving - situation clusters. - **Evaluate different models and methods**: Including supervised and unsupervised pre - trained visual - feature encoders, as well as multi - layer perceptrons (MLP) and situation - statistics - based driving - behavior predictors. ### Experimental results: - **Model performance**: The experimental results show that the proposed end - to - end training model and the clustering - based method are both superior to static driving - style models in predicting driving behaviors. - **Adaptability**: Especially in rural - road situations, the clustering - based method shows higher stability. - **Cluster specificity**: Through the ECS measure, it is verified that the identified driving - situation clusters have high specificity. In conclusion, this paper aims to improve the driving experience of autonomous driving systems and passengers' trust through the introduction of a situation - aware driving - style adaptation method.

Situation Awareness for Driver-Centric Driving Style Adaptation

Driving Behaviour Style Study with a Hybrid Deep Learning Framework Based on GPS Data

A Hybrid Deep Learning Approach for Driver Anomalous Lane Changing Identification

A Multimodal Data-Driven Approach for Driving Risk Assessment

A Scenario-Adaptive Driving Behavior Prediction Approach to Urban Autonomous Driving

Establishing Style-Oriented Driver Models by Imitating Human Driving Behaviors

A Study on Learning and Simulating Personalized Car-Following Driving Style

Human-like car-following modeling based on online driving style recognition

An Intention-aware and Online Driving Style Estimation Based Personalized Autonomous Driving Strategy

Self-Perception Versus Objective Driving Behavior: Subject Study of Lateral Vehicle Guidance

Multisource Adaption for Driver Attention Prediction in Arbitrary Driving Scenes

Towards Context-Aware Modeling of Situation Awareness in Conditionally Automated Driving

Driving Style Recognition of Interacting Vehicles in Multiple Scenarios

Using Eye-Tracking Data to Predict Situation Awareness in Real Time During Takeover Transitions in Conditionally Automated Driving

Human Insights Driven Latent Space for Different Driving Perspectives: A Unified Encoder for Efficient Multi-Task Inference

Constructing personalized situation awareness dataset for hazard perception, comprehension, projection, and action of drivers

Identifying Reaction-Aware Driving Styles of Stochastic Model Predictive Controlled Vehicles by Inverse Reinforcement Learning

Towards Driving Policies with Personality: Modeling Behavior and Style in Risky Scenarios via Data Collection in Virtual Reality

Multimodal Sensing and Computational Intelligence for Situation Awareness Classification in Autonomous Driving

Driver Intention Anticipation Based on In-Cabin and Driving Scene Monitoring

Adaptive Autopilot: Constrained DRL for Diverse Driving Behaviors