Application of Integrated Vehicle Health Management in Automated Decision-making for Driverless Vehicles
Time: Mon 2023-05-29 13.00
Location: F3, Lindstedtsvägen 26 & 28, Stockholm
Video link: https://kth-se.zoom.us/j/68692607321
Subject area: Machine Design
Doctoral student: Xin Tao , Mekatronik och inbyggda styrsystem, Integrated Transport Research Lab, ITRL
Opponent: Docent Tony Lindgren, Stockholms universitet
Supervisor: Professor Jonas Mårtensson, Integrated Transport Research Lab, ITRL, Reglerteknik
Vehicles are becoming increasingly complex and are prone to faults and failures, which threaten the dependability of vehicles in terms of availability, reliability, safety, and security. When vehicles are detected with certain types of faults and get into alarm situations, human drivers play a vital role in deciding what strategies and actions to take. Once driverless vehicles are introduced, human drivers' roles in decision-making will no longer exist, which urges new solutions on both technological and managerial levels.
This thesis depicts the current human decision-making process by analyzing field study data in the truck industry, which contributes to gaining domain knowledge and identifying research gaps. An integrated vehicle health management scheme is applied to automate this decision-making process by integrating vehicle health state estimation and prediction, resource utilization, and self-adaptive management. To implement this scheme, fault diagnosis and decision-making methods are proposed, and a decision support system is designed.
Fault diagnosis is a critical functional module for providing reliable vehicle health state information for decision-making. To address the influence of uncertainties in fault diagnosis, we propose an uncertainty analysis framework and a fault diagnosis method using Bayesian inference.Simulation experiments validate that the proposed method could effectively diagnose the root cause of fault symptoms under environmental uncertainty.
A risk-based automated decision-making method is presented, which imitates the human decision-making process.On this basis, a collaborative decision-making method is proposed by considering traffic congestion, which is a currently neglected public concern.Experiment results show that the proposed methods could effectively reduce the economic risk and the risk of traffic congestion.
In the end, a decision support system is designed to provide decision information to its human users. Besides, reviewing and learning functions are considered for gaining knowledge and achieving full automation in the long run. Additional system stakeholders from the public sector regarding safety, traffic, and the environment are considered. A transparent, interactive, and adaptive graphical user interface of the system is designed to enhance user experience and trust.
This thesis shows the potential of automated decision-making and technical system design in increasing corporate profits, catalyzing public-private partnerships, enabling technological transformation, and achieving a more sustainable transportation system.