Text2Scenario: Text-Driven Scenario Generation for Autonomous Driving Test

The denial-of-service (DoS) attack has evolved into a sophisticated and frequently utilized method of injecting the harmful files into the victim system in order to gain access to the sensitive and valuable data. This attack can cause a huge impact on various domains such as stock market exchange, banking sector, media sector, business websites, and e-commerce, and it will be very soon implemented in autonomous vehicles domain, which ultimately incurs huge national losses not only in terms of money but also in terms of the lives of people, when it is considered in the automotive sector especially in an autonomous vehicle. It has been observed that the DoS attack becomes very much popular as it gets the attention from all the researchers in the field of cybersecurity. In this paper, the comparative assessment of DoS attack methods on an autonomous vehicle test bed setup is proposed, and the analysis has been done in terms of different quality parameters such as time to attack, CPU utilization, memory utilization, packet loss rate, and average estimated round trip time. The attacker tool used here is Kali Linux. The results of the proposed experiments demonstrate which DoS attack approach is better for enhancing the performance of an autonomous car test bed configuration.KeywordsDoS attackTime to attackCPU utilizationMemory utilizationPacket loss rateRound trip timeAutonomous vehicle test bed setup

Autonomous driving techniques have been flourishing in recent years while thirsting for huge amounts of high-quality data. However, it is difficult for real-world datasets to keep up with the pace of changing requirements due to their expensive and time-consuming experimental and labeling costs. Therefore, more and more researchers are turning to synthetic datasets to easily generate rich and changeable data as an effective complement to the real world and to improve the performance of algorithms. In this paper, we summarize the evolution of synthetic dataset generation methods and review the work to date in synthetic datasets related to single and multi-task categories for the autonomous driving perception study. We also discuss the role that synthetic datasets play in the evaluation, gap test, and positive effect of autonomous driving-related algorithm testing, especially on trustworthiness and safety aspects, and provide examples of evaluation experiments. Finally, we discuss the limitations and future directions of synthetic datasets. To the best of our knowledge, this is the first survey focusing on the application of synthetic datasets in autonomous driving. This survey also raises awareness of the problems of real-world deployment of autonomous driving technology and provides researchers with a possible solution.

As autonomous driving technology matures, the focus shifts to enhancing the safety and reliability of these systems. Simulation testing is a critical method for efficiently and rapidly validating the performance of autonomous vehicles (AVs). A robust AV system requires extensive testing across a wide range of scenarios and iterative improvements. However, current simulation systems have limitations in supporting diverse scenarios, often relying on expert-designed situations. To address these challenges, we introduce DTTF-Sim, a novel simulation system based on Digital Twin technology for traffic flow. DTTF-Sim aims to accurately replicate real-world traffic flow conditions, offering continuous long-term simulation capabilities for AV testing. The system can simulate detailed dynamic traffic scenarios with a focus on interactions between multiple vehicles and between AVs and background traffic vehicles, modeling the strategic decision-making processes that occur in these encounters. This paper outlines the architecture and functionalities of DTTF-Sim, highlighting its ability to overcome the shortcomings of existing simulation platforms. We demonstrate the effectiveness of DTTF-Sim through case studies and experimental results, showing its potential to significantly advance the development and testing of autonomous driving technologies.

Objective Autonomous vehicles (AVs) have the potential to revolutionize the future of mobility by significantly improving traffic safety. This study presents a novel method for validating the safety performance of AVs in high-risk scenarios involving powered 2-wheelers (PTWs). By generating high-risk scenarios using in-depth crash data, this study is devoted to addressing the challenge of public road scenarios in testing, which often lack the necessary complexity and risk to effectively evaluate the capabilities of AVs in high-risk situations. Method Our approach employs a Wasserstein generative adversarial network (WGAN) to generate high-risk scenes, particularly focusing on PTW scenarios. By extracting 314 car-to-PTW crashes from the China In-depth Mobility Safety Study–Traffic Accident database, we simulate outcomes using PC-Crash software. The data are divided into scenes at 0.1-s intervals, with WGAN generating numerous high-risk scenes. By using a cumulative distribution function (CDF), we sampled and analyzed the vehicle’s dynamic information to generate complete scenarios applicable to the test. The validation process involves using the SVL Simulator and the Baidu Apollo joint simulation platform to evaluate the AV’s driving behavior and interactions with PTWs. Results This study evaluates model generation results by comparing distributions using Wasserstein distance as an indicator. The generator converges after approximately 200 epochs, with the iterator converging quickly. Subsequently, 10,000 new scenes are then generated. The distribution of several key parameters in the generated scenes can be found to approximate that of the original scenes. After sampling, the usability of generated scenarios is 64.76%. Virtual simulations confirm the effectiveness of the scenario generation method, with a generated scenario crash rate of 16.50% closely reflecting the original rate of 15.0%, showcasing the method’s capacity to produce realistic and hazardous scenarios. Conclusions The experimental results suggest that these scenarios exhibit a level of risk similar to the original crashes and are effective for testing AVs. Consequently, the generated scenarios enhance the diversity of the scenario library and accelerate the overall testing process of AVs.

Aiming at the problems of laser Simultaneous Localization and Mapping (SLAM), such as drift, low localization accuracy and map ghosting, which are prone to occur during localization and map construction in different structured environments, a tightly-coupled SLAM method of lidar and Inertial Measurement Unit (IMU) is proposed. Specifically, the initial position information provided by the IMU for the lidar is used to construct the joint error function of the IMU and the lidar, so as to achieve iterative position optimization. Following, the variety of the environmental structural information is determined, using a degradation detection method, based on the perturbation model, thus the selection of different methods for position estimation in different featured environments is achieved. Finally, when adequate number of structured features is obvious, the algorithm based online and surface feature matching, as proposed in LOAM, is used for motion estimation between neighboring frames. In the case of structural information lacking, the Iterative Closest Point (ICP) algorithm, containing distribution shape constraints, is introduced, to increase the constraints on the effective observation data, improving the stability and accuracy of the system, under the premise of ensuring that the system can run in real time. The results of extensive testing and validation in different scenarios, using open-source datasets, show that the proposed algorithm has higher accuracy and environmental adaptability, compared to other commonly used algorithms. In addition, real-vehicle tests, based on an autonomous self-driving car test platform, further demonstrate that the proposed method can be adapted to long-time drive and large-scale working environments, while it can operate steadily in real time.

The future of autonomous driving is fast approaching. As driving assistance and safety features increase, basic driving tasks will be automated, eliminating the requirement for human judgement and interaction. In today's world, features such as adaptive cruise control and lane departure warnings have become common. Using predictive ride technology, the car senses weather conditions and adjusts its ride settings to keep passengers comfortable on bumpy or uneven roads and to reduce travel sickness. This paper presents a thermal based vehicle detection system for autonomous car testing under extreme weather conditions. We utilize transfer learning by using three pre-trained convolutional neural networks and trained them on our data. The three networks were (GoogleNet, ResNet-50, SqueezeNet). The pre-trained networks showed different performance in terms of validation accuracy and frequency of alteration. The ResNet-50 network has the best performance, which shows a validation accuracy of 100%.

Traffic modeling is crucial for simulation tests of autonomous driving. High-fidelity and low-complexity traffic models facilitate traffic simulation for effective and efficient simulation test results. Recent works have shown the effectiveness of data-driven car-following models with various neural networks. In the pursuit of model accuracy, the complexity of the model increases. However, high-complexity models restrict large-scale traffic simulation, and the simulated car-following behavior may not guarantee long-term consistency with naturalistic driving data. Based on the analysis, we pointed out that the accumulation of model errors over time during simulation is the reason for the inconsistency. To deal with it, a loss-modified training method is proposed, to improve the long-term consistency without increasing the complexity of the model. Experiments conducted on real-world datasets demonstrate that the Loss Modified training method significantly improves the long-term consistency of car-following models and avoids abnormal speed during simulation.

Autonomous driving simulation tests play a crucial role in advancing autonomous driving technology and assessing its level of development. Conducting tests that incorporate traffic flow brings several advantages, including a comprehensive representation of various traffic elements and a close resemblance to real-world driving conditions. Such tests are effective in evaluating the performance of autonomous vehicles under different traffic scenarios. However, traditional traffic flow simulations primarily concentrate on enhancing the overall efficiency of the traffic system, rather than catering specifically to the testing requirements of autonomous vehicles within traffic flow. As a result, there is currently a shortage of comprehensive research summarizing the testing of autonomous vehicles in traffic flow environments. To address this issue, existing traffic models describing specific vehicle behavioral models and traffic flow-based autonomous driving test methods and metrics are first classified and summarized. In addition, future research on traffic flow-based autonomous driving tests is highlighted and discussed. By systematically organizing existing knowledge and highlighting potential avenues for research, this work aims to bridge the gap in understanding and facilitate the progress of autonomous driving testing within traffic flow environments.

Generating Autonomous Driving Hazard Test Scenarios Using Multi-Agent Proximal Policy Optimization and Enhanced Artificial Potential Field Method

Achieving the safety of autonomous vehicles (AVs) requires testing methodologies that are rigorous, scalable, and capable of addressing both routine and safety-critical situations. Traditional evaluation through mileage accumulation often fails to expose autonomous driving systems (ADS) to rare but high-impact events. Scenario-based testing has therefore emerged as a key paradigm, enabling the systematic construction of targeted and repeatable test cases. Central to this paradigm is scenario generation, which provides the essential means of populating simulations with realistic, diverse, and safety-relevant traffic situations. This survey presents a taxonomy and critical review of scenario generation methods for autonomous driving. We classify existing approaches into three paradigms—rule-based, data-driven, and learning-based—and analyze their methodologies, representative techniques, and supporting tools such as simulation platforms and scenario description languages. We also summarize evaluation metrics used to measure realism, diversity, and criticality, highlighting trade-offs between interpretability, scalability, and generalizability. Despite rapid progress, several challenges remain. Key issues include bridging the “reality gap” between virtual and real traffic, improving robustness under limited or biased datasets, and effectively modeling rare but safety-critical events. Emerging research directions are also explored, including language-driven generation, hybrid frameworks that integrate symbolic rules with generative learning, and standardized open scenario repositories to enhance reproducibility and regulatory acceptance. By consolidating diverse research efforts into a unified framework, this work provides structured insights for both researchers and practitioners. It aims to support the development of scalable and certifiable testing pipelines that are essential for the safe deployment of autonomous driving technologies.

Aim: This study aims to introduce search-based intelligent test generation for synthetic testing as a means of enabling optimal cross-layer testing in distributed software architectures comprising front-end interfaces, API gateways, microservices, databases, and cloud-based runtime services. It addresses the limitation of legacy quality assurance automation, which remains layer-isolated and rule-based, and is therefore unable to effectively detect inter-service faults, asynchronous failures, or unreliable execution behaviors in complex workflows. Methods: The research adopts a comparative mixed-methods approach, analyzing an AI-based framework that incorporates behavioral trace modeling, dependency graph generation, historical defect learning, and synthetic scenario generation. The evaluation was conducted across three types of enterprise applications, involving 30 autonomous test runs and 12 expert interviews. Results: The findings demonstrate substantial improvements in testing performance, with cross-layer validation coverage increasing to 84% and 62%, defects detected rising to 87% and 68%, and flaky tests reduced to 7% and 18%. Additionally, failure reproducibility improved by 19%, while scenario generation expanded by a factor of 2.7, increasing from 45 to 120 scenarios. Conclusion: The study concludes that synthetic test generation represents a realistic, scalable, and statistically informed approach that significantly enhances validation under realistic operating conditions, thereby providing a more reliable baseline for release confidence and operational decision-making. Recommendation: The study recommends further research focusing on reinforcement learning, explainability, increased autonomy through CI/CD integration, and the development of cross-domain benchmarks to advance the effectiveness and applicability of intelligent testing frameworks.

In this paper, we address the challenge of improving hazard detection in autonomous driving systems, particularly in scenarios where labeled data is scarce or unavailable. This issue is critical in real-world applications, where diverse and unpre- dictable driving situations make it difficult to label every poten- tial hazard accurately. Recently, the Challenge of Out-of-Label (COOOL) benchmark has been introduced at WACV2025 to promote research on this challenge.To tackle this issue, we present a novel method that integrates a Bootstrapping Language-Image Pretraining (BLIP)-based scenario generation framework with a threshold-based hazard scoring system, thereby enhancing both scenario comprehension and detection accuracy within the benchmark. By incorporating robust driver state logic, bounding box analysis, and BLIP-generated scenario descriptions, our method initially achieves a 40% performance score. Building upon this foundation, we further integrate depth maps and optical flow to improve hazardous object discrimination, resulting in an additional 20% performance improvement. This culminates in a final score of 63% on the public benchmark leaderboard and 50% on the private leaderboard. To foster continued ad- vancements in autonomous driving research, we will make all code and visualization tools publicly available.