This article presents a novel framework for modeling the troubleshooting process for automotive systems such as trucks and buses. We describe how a diagnostic model of the troubleshooting process can be created using event-driven, nonstationary, dynamic Bayesian networks. Exact inference in such a model is in general not practically possible. Therefore, we evaluate different approximate methods for inference based on the Boyen–Koller algorithm. We identify relevant model classes that have particular structure such that inference can be made with linear time complexity. We also show how models created using expert knowledge can be tuned using statistical data. The proposed learning mechanism can use data that is collected from a heterogeneous fleet of modular vehicles that can consist of different components. The proposed framework is evaluated both theoretically and experimentally on an application example of a fuel injection system.

This thesis considers computer-assisted troubleshooting of heavy vehicles such as trucks and buses. In this setting, the person that is troubleshooting a vehicle problem is assisted by a computer that is capable of listing possible faults that can explain the problem and gives recommendations of which actions to take in order to solve the problem such that the expected cost of restoring the vehicle is low. To achieve this, such a system must be capable of solving two problems: the diagnosis problem of finding which the possible faults are and the decision problem of deciding which action should be taken.

The diagnosis problem has been approached using Bayesian network models. Frameworks have been developed for the case when the vehicle is in the workshop only and for remote diagnosis when the vehicle is monitored during longer periods of time.

The decision problem has been solved by creating planners that select actions such that the expected cost of repairing the vehicle is minimized. New methods, algorithms, and models have been developed for improving the performance of the planner.

The theory developed has been evaluated on models of an auxiliary braking system, a fuel injection system, and an engine temperature control and monitoring system.

When a truck or bus suffers from a breakdown it is important that the vehicle comes back on the road as soon as possible. In this paper we present a prototype diagnostic decision support system capable of automatically identifying possible causes of a failure and propose recommended actions on how to get the vehicle back on the road as cost efficiently as possible.

This troubleshooting system is novel in the way it integrates the remote diagnosis with the workshop diagnosis when providing recommendations. To achieve this integration, a novel planning algorithm has been developed that enables the troubleshooting system to guide the different users (driver, help-desk operator, and mechanic) through the entire troubleshooting process.

In this paper we formulate the problem of integrated remote and workshop troubleshooting and present a working prototype that has been implemented to demonstrate all parts of the troubleshooting system.

When parts of the states in a goal POMDP are fully observable and some actions are deterministic it is possibleto take advantage of these properties to efficiently generate approximate solutions. Actions that deterministically affect the fully observable component of the world state can be abstracted away and combined into macro actions, permitting a planner to converge more quickly. This processing can be separated from the main search procedure, allowing us to leverage existing POMDP solvers. Theoretical results show how a POMDP can be analyzed to identify the exploitable properties and formal guarantees are provided showing that the use of macro actions preserves solvability. The efficiency of the method is demonstrated with examples when used in combination with existing POMDP solvers.

Computer assisted troubleshooting with external interventions is considered. The work is motivated by the task of repairing an automotive vehicle at lowest possible expected cost. The main contribution is a decision theoretic troubleshooting system that is developed to handle external interventions. In particular, practical issues in modeling for troubleshooting are discussed, the troubleshooting system is described, and a method for the efficient probability computations is developed. The troubleshooting systems consists of two parts; a planner that relies on AO* search and a diagnoser that utilizes Bayesian networks (BN). The work is based on a case study of an auxiliary braking system of a modern truck. Two main challenges in troubleshooting automotive vehicles are the need for disassembling the vehicle during troubleshooting to access parts to repair, and the difficulty to verify that the vehicle is fault free. These facts lead to that probabilities for faults and for future observations must be computed for a system that has been subject to external interventions that cause changes in the dependency structure. The probability computations are further complicated due to the mixture of instantaneous and non-instantaneous dependencies. To compute the probabilities, we develop a method based on an algorithm, updateBN, that updates a static BN to account for the external interventions.

This licentiate thesis considers computer-assisted troubleshooting of complex products such as heavy trucks. The troubleshooting task is to find and repair all faulty components in a malfunctioning system. This is done by performing actions to gather more information regarding which faults there can be or to repair components that are suspected to be faulty. The expected cost of the performed actions should be as low as possible.

The work described in this thesis contributes to solving the troubleshooting task in such a way that a good trade-off between computation time and solution quality can be made. A framework for troubleshooting is developed where the system is diagnosed using non-stationary dynamic Bayesian networks and the decisions of which actions to perform are made using a new planning algorithm for Stochastic Shortest Path Problems called Iterative Bounding LAO*.

It is shown how the troubleshooting problem can be converted into a Stochastic Shortest Path problem so that it can be efficiently solved using general algorithms such as Iterative Bounding LAO*.  New and improved search heuristics for solving the troubleshooting problem by searching are also presented in this thesis.

The methods presented in this thesis are evaluated in a case study of an auxiliary hydraulic braking system of a modern truck. The evaluation shows that the new algorithm Iterative Bounding LAO* creates troubleshooting plans with a lower expected cost faster than existing state-of-the-art algorithms in the literature. The case study shows that the troubleshooting framework can be applied to systems from the heavy vehicles domain.

Iterative Bounding LAO* is a new algorithm for epsilon- optimal probabilistic planning problems where an absorbing goal state should be reached at a minimum expected cost from a given initial state. The algorithm is based on the LAO* algorithm for finding optimal solutions in cyclic AND/OR graphs. The new algorithm uses two heuristics, one upper bound and one lower bound of the optimal cost. The search is guided by the lower bound as in LAO*, while the upper bound is used to prune search branches. The algorithm has a new mechanism for expanding search nodes, and while maintaining the error bounds, it may use weighted heuristics to reduce the size of the explored search space. In empirical tests on benchmark problems, Iterative Bounding LAO* expands fewer search nodes compared to state of the art RTDP variants that also use two-sided bounds.

We consider computer assisted troubleshooting of complex systems, for example of a vehicle at a workshop. The objective is to identify the cause of a failure and repair a system at as low expected cost as possible. Three main challenges are: the need for disassembling the system during troubleshooting, the difficulty to verify that the system is fault free, and the dependencies in between components and observations. We present a method that can return a response anytime, which allows us to obtain the best result given the available time. The work is based on a case study of an auxiliary braking system of a modern truck. We highlight practical issues related to model building and troubleshooting in a real environment.

We consider computer assisted troubleshooting of automotive vehicles, where the objective is to repair the vehicle at as low expected cost as possible.

The work has three main contributions: a troubleshooting method that applies to troubleshooting in real environments, the discussion on practical issues in modeling for troubleshooting, and the efficient probability computations.

The work is based on a case study of an auxiliary braking system of a modern truck.

We apply a decision theoretic approach, consisting of a planner and a diagnoser.

Two main challenges in troubleshooting automotive vehicles are the need for disassembling the vehicle during troubleshooting to access parts to repair, and the difficulty to verify that the vehicle is fault free. These facts lead to that probabilities for faults and for future observations must be computed for a system that has been subject to external interventions that cause changes the dependency structure. The probability computations are further complicated due to the mixture of instantaneous and non-instantaneous dependencies.

To compute the probabilities, we develop a method based on an algorithm, updateBN, that updates a static BN to account for the external interventions.

We consider computer assisted troubleshooting of complex systems, where the objective is to identify the cause of a failure and repair the system at as low expected cost as possible. Three main challenges are: the need for disassembling the system during troubleshooting, the difficulty to verify that the system is fault free, and the dependencies in between components and observations. We present a method that can return a response anytime, which allows us to obtain the best result given the available time. The work is based on a case study of an auxiliary braking system of a modern truck. We highlight practical issues related to model building and troubleshooting in a real environment.