Abstract:Evaluating the decision-making system is indispensable in developing autonomous vehicles, while realistic and challenging safety-critical test scenarios play a crucial role. Obtaining these scenarios is non-trivial, thanks to the long-tailed distribution, sparsity, and rarity in real-world data sets. To tackle this problem, in this paper, we introduce a natural adversarial scenario generation solution using naturalistic human driving priors and reinforcement learning techniques. By doing this, we can obtain large-scale test scenarios that are both diverse and realistic. Specifically, we build a simulation environment that mimics natural traffic interaction scenarios. Informed by this environment, we implement a two-stage procedure. The first stage incorporates conventional rule-based models, e.g., IDM~(Intelligent Driver Model) and MOBIL~(Minimizing Overall Braking Induced by Lane changes) model, to coarsely and discretely capture and calibrate key control parameters from the real-world dataset. Next, we leverage GAIL~(Generative Adversarial Imitation Learning) to represent driver behaviors continuously. The derived GAIL can be further used to design a PPO~(Proximal Policy Optimization)-based actor-critic network framework to fine-tune the reward function, and then optimizes our natural adversarial scenario generation solution. Extensive experiments have been conducted in the NGSIM dataset including the trajectory of 3,000 vehicles. Essential traffic parameters were measured in comparison with the baseline model, e.g., the collision rate, accelerations, steering, and the number of lane changes. Our findings demonstrate that the proposed model can generate realistic safety-critical test scenarios covering both naturalness and adversariality, which can be a cornerstone for the development of autonomous vehicles.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to generate safety - critical test scenarios that are both challenging and realistic in the evaluation of decision - making systems in autonomous vehicles (AVs). Specifically, the paper points out that in real - world datasets, these scenarios are difficult to obtain due to their long - tailed distribution, sparsity, and rarity. Therefore, the paper proposes an adversarial safety - critical scenario generation solution based on natural human driving priors and reinforcement learning techniques, aiming to generate large - scale and diverse test scenarios that can not only reflect real - world traffic conditions but also contain adversarial elements to challenge the limits of autonomous driving systems. To achieve this goal, the paper first constructs an environment that simulates natural traffic interaction scenarios and implements a two - stage procedure on this basis. The first stage uses traditional rule - based models (such as IDM and MOBIL models) to roughly capture and calibrate key control parameters from real - world datasets; the second stage uses Generative Adversarial Imitation Learning (GAIL) to continuously represent driver behavior, and then designs an actor - critic network framework based on Proximal Policy Optimization (PPO) to fine - tune the reward function and optimize the natural adversarial scenario generation solution. Through extensive experiments on two popular datasets (NGSIM and INTERACTION), the paper shows that the proposed model can generate highly adversarial safety - critical test scenarios while maintaining naturalness, with an efficiency improvement of 44% compared to the baseline model. This provides an important test basis for the development of autonomous vehicles.

Adversarial Safety-Critical Scenario Generation using Naturalistic Human Driving Priors

Adversarial Generation of Safety-Critical Lane-Change Scenarios for Autonomous Vehicles

Game-Theoretic Adversarial Interaction-Based Critical Scenario Generation for Autonomous Vehicles

Interactive Critical Scenario Generation for Autonomous Vehicles Testing Based on In-depth Crash Data Using Reinforcement Learning

AdvSim: Generating Safety-Critical Scenarios for Self-Driving Vehicles

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

A Survey on Safety-Critical Driving Scenario Generation—A Methodological Perspective

Critical Scenario Generation for Realistic Testing of Autonomous Driving System Via Multi-vehicle Cooperative Control

ReGentS: Real-World Safety-Critical Driving Scenario Generation Made Stable

Learning Naturalistic Driving Environment with Statistical Realism

Adversarial Driving Behavior Generation Incorporating Human Risk Cognition for Autonomous Vehicle Evaluation

DeepMF: Deep Motion Factorization for Closed-Loop Safety-Critical Driving Scenario Simulation

Double Counterfactual Regret Minimization for Generating Safety-Critical Scenario of Autonomous Driving

Quantitative Representation of Scenario Difficulty for Autonomous Driving Based on Adversarial Policy Search

Controllable Diffusion Models for Safety-Critical Driving Scenario Generation

Efficient Generation of Safety-Critical Scenarios Combining Dynamic and Static Scenario Parameters

Multimodal Safety-Critical Scenarios Generation for Decision-Making Algorithms Evaluation

Scenario Generation for Autonomous Vehicles with Deep-Learning-Based Heterogeneous Driver Models: Implementation and Verification

Risk Scenario Generation for Autonomous Driving Systems based on Causal Bayesian Networks

A Dynamic Test Scenario Generation Method for Autonomous Vehicles Based on Conditional Generative Adversarial Imitation Learning

Automated and Complete Generation of Traffic Scenarios at Road Junctions Using a Multi-level Danger Definition