Abstract:In autonomous vehicle (AV) technology, the ability to accurately predict the movements of surrounding vehicles is paramount for ensuring safety and operational efficiency. Incorporating human decision-making insights enables AVs to more effectively anticipate the potential actions of other vehicles, significantly improving prediction accuracy and responsiveness in dynamic environments. This paper introduces the Human-Like Trajectory Prediction (HLTP) model, which adopts a teacher-student knowledge distillation framework inspired by human cognitive processes. The HLTP model incorporates a sophisticated teacher-student knowledge distillation framework. The "teacher" model, equipped with an adaptive visual sector, mimics the visual processing of the human brain, particularly the functions of the occipital and temporal lobes. The "student" model focuses on real-time interaction and decision-making, drawing parallels to prefrontal and parietal cortex functions. This approach allows for dynamic adaptation to changing driving scenarios, capturing essential perceptual cues for accurate prediction. Evaluated using the Macao Connected and Autonomous Driving (MoCAD) dataset, along with the NGSIM and HighD benchmarks, HLTP demonstrates superior performance compared to existing models, particularly in challenging environments with incomplete data. The project page is available at Github.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve a core problem in autonomous driving technology: how to accurately predict the driving trajectories of surrounding vehicles. Specifically, the paper points out that although existing deep - learning models have made significant progress in handling vehicle pattern recognition and environmental data, there is still an obvious gap in simulating complex human cognitive functions. Current models can handle structured data, but perform poorly in scenarios that require the adaptability and predictability of human drivers. To bridge this gap, the paper proposes a model named Human - Like Trajectory Prediction (HLTP), which adopts an approach based on the teacher - student knowledge distillation framework to simulate the human cognitive process. ### Main contributions 1. **Introduced a novel visual perception pooling mechanism**: The HLTP model introduced an adaptive visual area, which can be dynamically adjusted according to the vehicle speed, so as to more effectively capture key perception cues in different traffic scenarios. 2. **Adopted a cognitive - inspired teacher - student knowledge distillation framework**: By efficiently integrating complex visual attention and spatial awareness patterns, the prediction accuracy and robustness are improved, especially in complex scenarios. 3. **Introduced a new Macau Connected Autonomous Driving (MoCAD) dataset**: This dataset provides a new urban driving environment for right - hand - drive systems, providing a new context for trajectory prediction research. ### Method overview - **Problem definition**: The goal of the paper is to predict the trajectories of all vehicles around an autonomous vehicle in a mixed - autonomy environment. Given the historical trajectory data of the target vehicle and its surrounding vehicles, the model aims to predict the future multi - modal probability distribution of the target vehicle. - **Model architecture**: The HLTP model consists of a "teacher" model and a "student" model. Among them, the "teacher" model simulates the human visual observation process, using the Shift - Window Attention block (SWA) and the peripheral perception encoder to capture the information of central vision and peripheral vision. The "student" model adopts a lightweight framework and learns from the "teacher" model through knowledge distillation to improve the prediction accuracy. - **Scene representation**: HLTP describes traffic scenes through information such as relative position, speed, and acceleration, and converts them into visual vectors and context matrices for further feature extraction. - **Visual perception pooling mechanism**: This mechanism simulates the visual attention distribution of human drivers, and more precisely captures important details by introducing a visual weight matrix to dynamically adjust the field of view. - **Training method**: The paper details the training methods of the "teacher" model and the "student" model, including multi - task learning and the design of loss functions, to ensure the accuracy and robustness of the model. ### Conclusion By introducing the adaptive visual area and the cognitive - inspired knowledge distillation framework, the HLTP model performs well in trajectory prediction, especially in scenarios with incomplete data or complex environments. This model is not only innovative in technology, but also deeply explores the cognitive mechanisms of the human brain, providing new ideas for improving the decision - making ability of autonomous vehicles.

A Cognitive-Based Trajectory Prediction Approach for Autonomous Driving

Less is More: Efficient Brain-Inspired Learning for Autonomous Driving Trajectory Prediction

A Cognitive-Driven Trajectory Prediction Model for Autonomous Driving in Mixed Autonomy Environment

Human Observation-Inspired Trajectory Prediction for Autonomous Driving in Mixed-Autonomy Traffic Environments

Enhanced Multimodal Trajectory Prediction for Autonomous Vehicles Using Advanced Diffusion Model Techniques

Self-Aware Trajectory Prediction for Safe Autonomous Driving

MFTraj: Map-Free, Behavior-Driven Trajectory Prediction for Autonomous Driving

BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving

Probabilistic multi-modal expected trajectory prediction based on LSTM for autonomous driving

A Cognition‐inspired Trajectory Prediction Method for Vehicles in Interactive Scenarios

Characterized Diffusion and Spatial-Temporal Interaction Network for Trajectory Prediction in Autonomous Driving

A Probabilistic Architecture of Long-Term Vehicle Trajectory Prediction for Autonomous Driving

Enabling Accurate Trajectory Prediction of Human Driven Vehicles in the Hybrid Driving Scenario

Long Short-Term Memory-Based Human-Driven Vehicle Longitudinal Trajectory Prediction in a Connected and Autonomous Vehicle Environment

Context‐aware trajectory prediction for autonomous driving in heterogeneous environments

AI-TP: Attention-based Interaction-aware Trajectory Prediction for Autonomous Driving

Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline

Intention-based and Risk-Aware Trajectory Prediction for Autonomous Driving in Complex Traffic Scenarios

Probabilistic Long-term Vehicle Trajectory Prediction Via Driver Awareness Model

Vehicle Trajectory Prediction based Predictive Collision Risk Assessment for Autonomous Driving in Highway Scenarios