Failure Prediction for Autonomous Driving

Simon Hecker,Dengxin Dai,Luc Van Gool

DOI: https://doi.org/10.48550/arXiv.1805.01811

2018-05-04

Abstract:The primary focus of autonomous driving research is to improve driving accuracy. While great progress has been made, state-of-the-art algorithms still fail at times. Such failures may have catastrophic consequences. It therefore is important that automated cars foresee problems ahead as early as possible. This is also of paramount importance if the driver will be asked to take over. We conjecture that failures do not occur randomly. For instance, driving models may fail more likely at places with heavy traffic, at complex intersections, and/or under adverse weather/illumination conditions. This work presents a method to learn to predict the occurrence of these failures, i.e. to assess how difficult a scene is to a given driving model and to possibly give the human driver an early headsup. A camera-based driving model is developed and trained over real driving datasets. The discrepancies between the model's predictions and the human `ground-truth' maneuvers were then recorded, to yield the `failure' scores. Experimental results show that the failure score can indeed be learned and predicted. Thus, our prediction method is able to improve the overall safety of an automated driving model by alerting the human driver timely, leading to better human-vehicle collaborative driving.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is, in autonomous driving, how to predict the situations where the driving model may fail, so as to issue warnings to human drivers in advance, thereby improving the overall safety of the autonomous driving system. Specifically, the paper focuses on developing a method to learn and predict the occurrence of these failures, that is, to evaluate the difficulty of a given driving model in specific scenarios, and possibly notify human drivers to take over the vehicle in advance. Such prediction is crucial for ensuring the safety and reliability of autonomous driving, especially in complex traffic environments, bad weather or lighting conditions where the driving model is more likely to make mistakes. The paper develops a camera - based driving model and uses a real - driving data set for training, records the differences between the model predictions and human "ground truth" operations, and generates "failure" scores. The experimental results show that the failure scores can indeed be learned and predicted, so that the overall safety of the automated driving model can be improved by timely reminding human drivers, promoting better human - vehicle collaborative driving.

Failure Prediction for Autonomous Driving

Failure Detection for Motion Prediction of Autonomous Driving: An Uncertainty Perspective

Prediction Failure Risk-Aware Decision-Making for Autonomous Vehicles on Signalized Intersections

Improving Interaction-based Vehicle Trajectory Prediction Via Handling Sensing Failures

Learning Accurate, Comfortable and Human-like Driving

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

Adaptive Visual Interaction Based Multi-Target Future State Prediction For Autonomous Driving Vehicles

Self-Aware Trajectory Prediction for Safe Autonomous Driving

When, Where and How Does It Fail? A Spatial-Temporal Visual Analytics Approach for Interpretable Object Detection in Autonomous Driving.

Failure Prediction in Vision Systems

Is it Safe to Drive? An Overview of Factors, Challenges, and Datasets for Driveability Assessment in Autonomous Driving

Infrastructure-based End-to-End Learning and Prevention of Driver Failure

A Review of Deep Learning-Based Vehicle Motion Prediction for Autonomous Driving

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

Predicting Driver Errors during Automated Vehicle Takeovers

Machine Learning-Based Vehicle Intention Trajectory Recognition and Prediction for Autonomous Driving

Driver Behavior Failure Probability Prediction Based on CREAM

Modelling of Longitudinal and Lateral Behavior of Drivers and Automatic Prediction-evaluation

Surrogate Evaluation Model for Assessing Lane Detection Reliability of Automated Vehicles in Complex Road Environments

Machine Learning for Prediction of Driver Takeover Time in Automated Driving: Insights from Non-urgent Low Consequence Scenarios

Is it Safe to Drive? An Overview of Factors, Metrics, and Datasets for Driveability Assessment in Autonomous Driving