Abstract:Hybrid traffic which involves both autonomous and human-driven vehicles would be the norm of the autonomous vehicles practice for a while. On the one hand, unlike autonomous vehicles, human-driven vehicles could exhibit sudden abnormal behaviors such as unpredictably switching to dangerous driving modes, putting its neighboring vehicles under risks; such undesired mode switching could arise from numbers of human driver factors, including fatigue, drunkenness, distraction, aggressiveness, etc. On the other hand, modern vehicle-to-vehicle communication technologies enable the autonomous vehicles to efficiently and reliably share the scarce run-time information with each other. In this paper, we propose, to the best of our knowledge, the first efficient algorithm that can (1) significantly improve trajectory prediction by effectively fusing the run-time information shared by surrounding autonomous vehicles, and can (2) accurately and quickly detect abnormal human driving mode switches or abnormal driving behavior with formal assurance without hurting human drivers privacy. To validate our proposed algorithm, we first evaluate our proposed trajectory predictor on NGSIM and Argoverse datasets and show that our proposed predictor outperforms the baseline methods. Then through extensive experiments on SUMO simulator, we show that our proposed algorithm has great detection performance in both highway and urban traffic. The best performance achieves detection rate of 97.3%, average detection delay of 1.2s, and 0 false alarm.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to achieve the safe autonomy of self - driving vehicles in mixed traffic. Specifically, the paper focuses on how to detect the unpredictable abnormal behaviors of human drivers, which may occur suddenly, such as switching to a dangerous driving mode, thus causing potential safety hazards to surrounding vehicles. The root causes of these problems may be factors such as fatigue, drunk driving, distraction or aggression. The paper proposes a new and efficient algorithm, which can significantly improve the accuracy of trajectory prediction by effectively fusing the real - time information shared by surrounding self - driving vehicles, and can quickly and accurately detect the switching of abnormal driving modes without violating the privacy of human drivers. To verify the proposed algorithm, the authors first evaluated their trajectory predictor on the NGSIM and Argoverse datasets, and the results showed that its performance was better than the baseline methods. Then, through extensive experiments on the SUMO simulator, it was proved that the algorithm had excellent detection performance in highway and urban traffic. The best performance achieved a detection rate of 97.3%, with an average detection delay of 1.2 seconds and no false alarms. The contributions of the paper are mainly reflected in two aspects: 1. **Trajectory Prediction**: A trajectory prediction model based on the multi - encoder attention mechanism (MEATP) is proposed, which significantly improves the accuracy of trajectory prediction by using the information shared between vehicles. 2. **Anomaly Detection**: The CuSum algorithm and its variants are used to monitor the pattern changes of the prediction error at runtime, so as to accurately and quickly detect the occurrence of abnormal driving behaviors. The algorithm considers two scenarios: the situation where the probability distributions before and after the change are fully known, and the situation where only the post - change distribution is partially known. Experiments show that the algorithm performs well in different scenarios and is robust to observation noise. In general, the paper proposes an innovative method aiming to improve the safety and reliability of self - driving vehicles in a mixed traffic environment through vehicle - to - vehicle communication technologies.

Towards Safe Autonomy in Hybrid Traffic: Detecting Unpredictable Abnormal Behaviors of Human Drivers via Information Sharing

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

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

Abnormal Driving Detection Based on Normalized Driving Behavior

Human Driver Behavior Prediction based on UrbanFlow

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

Autonomous Driving Simulation Platform for Hybrid Traffic

A Hybrid Trajectory Prediction Framework for Automated Vehicles with Attention Mechanisms

Self-Aware Trajectory Prediction for Safe Autonomous Driving

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

Trajectory Prediction for Human-driven Vehicles Based on a Combined Method of Deep Neural Network and Driving Risk Map

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

Transferable and Adaptable Driving Behavior Prediction

Traffic Conflict Forecasting and Avoidance System Under Automated Driving System Disengagement: A Non-Intrusive Prototype Design

Towards Safe Motion Planning for Autonomous Driving in Highway

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

Safe, Efficient and Socially-Compatible Decision of Automated Vehicles: a Case Study of Unsignalized Intersection Driving

Probabilistic Vehicle Trajectory Prediction Via Driver Characteristic and Intention Estimation Model under Uncertainty

Driver Intent-Based Intersection Autonomous Driving Collision Avoidance Reinforcement Learning Algorithm

A Method for Predicting Diverse Lane-Changing Trajectories of Surrounding Vehicles Based on Early Detection of Lane Change.

Prediction-Based Reachability for Collision Avoidance in Autonomous Driving