Abstract:Recent advances in Deep Neural Networks (DNNs) have led to the development of DNN-driven autonomous cars that, using sensors like camera, LiDAR, etc., can drive without any human intervention. Most major manufacturers including Tesla, GM, Ford, BMW, and Waymo/Google are working on building and testing different types of autonomous vehicles. The lawmakers of several US states including California, Texas, and New York have passed new legislation to fast-track the process of testing and deployment of autonomous vehicles on their roads. However, despite their spectacular progress, DNNs, just like traditional software, often demonstrate incorrect or unexpected corner case behaviors that can lead to potentially fatal collisions. Several such real-world accidents involving autonomous cars have already happened including one which resulted in a fatality. Most existing testing techniques for DNN-driven vehicles are heavily dependent on the manual collection of test data under different driving conditions which become prohibitively expensive as the number of test conditions increases. In this paper, we design, implement and evaluate DeepTest, a systematic testing tool for automatically detecting erroneous behaviors of DNN-driven vehicles that can potentially lead to fatal crashes. First, our tool is designed to automatically generated test cases leveraging real-world changes in driving conditions like rain, fog, lighting conditions, etc. DeepTest systematically explores different parts of the DNN logic by generating test inputs that maximize the numbers of activated neurons. DeepTest found thousands of erroneous behaviors under different realistic driving conditions (e.g., blurring, rain, fog, etc.) many of which lead to potentially fatal crashes in three top performing DNNs in the Udacity self-driving car challenge.

SDLV: Verification of Steering Angle Safety for Self-Driving Cars

A Deep Learning Method for Lane Changing Situation Assessment and Decision Making.

A Verification Framework for Behavioral Safety of Self‐driving Cars

Are Self-Driving Cars Secure? Evasion Attacks against Deep Neural Networks for Steering Angle Prediction

A Vision-Based Method for Improving the Safety of Self-Driving

Deep Steering: Learning End-to-End Driving Model from Spatial and Temporal Visual Cues

Advancing Autonomous Navigation: Deep Learning Techniques for Self-Driving Cars

Vision-based Navigation of Autonomous Vehicle in Roadway Environments with Unexpected Hazards

Deep learning approach for accurate and stable recognition of driver's lateral intentions using naturalistic driving data

DeepTest: Automated Testing of Deep-Neural-Network-driven Autonomous Cars

Evaluating Adversarial Attacks on Driving Safety in Vision-Based Autonomous Vehicles

Deep Learning for Safe Autonomous Driving: Current Challenges and Future Directions

Towards Safety Verification of Direct Perception Neural Networks

Deep Learning Safety Concerns in Automated Driving Perception

Learning End-to-End Autonomous Steering Model from Spatial and Temporal Visual Cues

Towards Safety Assured End-to-End Vision-Based Control for Autonomous Racing

Dense reinforcement learning for safety validation of autonomous vehicles

Deep Learning Serves Traffic Safety Analysis: A Forward-looking Review

Reliability of Deep Neural Networks for an End-to-End Imitation Learning-Based Lane Keeping

Verified Safe Reinforcement Learning for Neural Network Dynamic Models

Runtime Safety Assurance for Learning-enabled Control of Autonomous Driving Vehicles