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

Haicheng Liao,Zhenning Li,Chengyue Wang,Bonan Wang,Hanlin Kong,Yanchen Guan,Guofa Li,Zhiyong Cui,Chengzhong Xu
2024-04-27
Abstract:As autonomous driving technology progresses, the need for precise trajectory prediction models becomes paramount. This paper introduces an innovative model that infuses cognitive insights into trajectory prediction, focusing on perceived safety and dynamic decision-making. Distinct from traditional approaches, our model excels in analyzing interactions and behavior patterns in mixed autonomy traffic scenarios. It represents a significant leap forward, achieving marked performance improvements on several key datasets. Specifically, it surpasses existing benchmarks with gains of 16.2% on the Next Generation Simulation (NGSIM), 27.4% on the Highway Drone (HighD), and 19.8% on the Macao Connected Autonomous Driving (MoCAD) dataset. Our proposed model shows exceptional proficiency in handling corner cases, essential for real-world applications. Moreover, its robustness is evident in scenarios with missing or limited data, outperforming most of the state-of-the-art baselines. This adaptability and resilience position our model as a viable tool for real-world autonomous driving systems, heralding a new standard in vehicle trajectory prediction for enhanced safety and efficiency.
Robotics
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to predict vehicle trajectories more accurately in a mixed autonomous driving environment. Specifically, the paper focuses on how to improve the prediction accuracy of vehicle trajectories by introducing a cognition - driven method in the situation where autonomous vehicles (AVs) coexist with human - driven vehicles (HVs). Traditional methods have limitations in dealing with such complex interactions, especially in understanding and predicting human driving behaviors. Therefore, the paper proposes a new trajectory prediction model, aiming to improve dynamic decision - making and behavior pattern analysis by integrating cognitive insights, especially the concept of perceived safety, so as to achieve significant performance improvements on multiple key datasets. The key contributions of the paper include: 1. **Development of the Perceived Safety Awareness Module**: This module contains two core parts, namely Quantitative Safety Assessment (QSA) and Driver Behavior Profiling (DBP). QSA objectively assesses the safety of driving scenarios, while DBP adds a human element on this basis to distinguish the behavioral characteristics of different drivers. 2. **Development of the Social Interaction Awareness Module**: A Transformer - based framework (called Leanformer) is designed to capture the complex interactions between vehicles. 3. **Robustness to Data Inconsistency**: The model performs well in the face of incomplete or inconsistent data, ensuring accurate and reliable trajectory prediction even when the data is challenging in real - world applications. Through these innovations, the paper aims to provide an autonomous driving system with a prediction method that is closer to human driving behaviors, thereby improving road safety and efficiency.