Papers by Martin Sachenbacher
Lecture Notes in Computer Science, 2009
Testing is the process of stimulating a system with inputs in order to reveal hidden parts of the... more Testing is the process of stimulating a system with inputs in order to reveal hidden parts of the system state. In the case of nondeterministic systems, the difficulty arises that an input pattern can generate several possible outcomes. Some of these outcomes allow to distinguish between different hypotheses about the system state, while others do not. In this paper, we present a novel approach to find, for non-deterministic systems modeled as constraints over variables, tests that allow to distinguish among the hypotheses as good as possible. The idea is to assess the quality of a test by determining the ratio of distinguishing (good) and not distinguishing (bad) outcomes. This measure refines previous notions proposed in the literature on model-based testing and can be computed using model counting techniques. We propose and analyze a greedy-type algorithm to solve this test optimization problem, using existing model counters as a building block. We give preliminary experimental results of our method, and discuss possible improvements.
Conflict-driven problem solvers such as GDE use previously discovered conflicts to guide further ... more Conflict-driven problem solvers such as GDE use previously discovered conflicts to guide further search through the candidate space. To do so, ATMS-based problem solvers employ an inference engine that performs two fundamentally different tasks: Checking a given assumption set for consistency and predicting values for system variables under given assumptions. In this paper, we show how separating the tasks of conflict search and prediction of values leads to a problem solver that can guarantee completeness and correctness of the consistency check for a given assumption set, and how this can be used to effectively guide the search for minimal conflicts. We develop an inference engine based on aggregation of relational models that can be shown to have the above properties, provided that three basic operations on relations are available. As a consequence, we can complement conflict-driven diagnosis by a new paradigm called consistency-driven search for conflicts. To illustrate these po...
AI Communications, 2000
Within the European "Vehicle Model based Diagnosis" (VMBD) project, demonstrator vehicles with bu... more Within the European "Vehicle Model based Diagnosis" (VMBD) project, demonstrator vehicles with built-in faults provided a serious challenge to model-based diagnosis techniques and a real-life test-bed for their evaluation. One of the guiding applications within VMBD was model-based on-board diagnosis of faults in a turbo diesel engine system with a focus on potential origins of increased carbon emissions. Our goal was to build a demonstrator system that is able to run on-board, processing the signals available to the ordinary control unit. This paper does not aim at presenting new theories and technologies, but focuses on the application aspects and the lessons learned: What did work? And what needs to be done? First, we analyze the requirements imposed, the way they were addressed by the chosen solutions, and the results obtained by the on-board diagnosis prototype running on the demonstrator vehicle. The most important challenges of the demonstrator were to apply model-based diagnosis systems to dynamic systems with feedback, to handle systems without a rigorous mathematical model (such as the combustion engine), and to try to provide the response times required for real-time applications. Second, we discuss, based on our experience, the obstacles to transferring the technology to industrial application that we found or foresee. The main ones are related to modeling and a narrow perspective on diagnostic processes; some of the problems are caused by the inertia of current work process organization and the fact that modeling and diagnosis in industry is the domain of engineers, who maintain views and techniques that are rather different from ours. Despite these non-technical origins, the problems raised pose demanding challenges to research in modelbased systems, and we believe they might be helpful for the research community to guide and focus its efforts. Research on model-based diagnosis (e.g. [Hamscher et al. 92], [Dressler and Struss 96]) has generated a number of well-founded theories and sophisticated prototypes of implemented diagnosis engines. However, many of these
With robots becoming increasingly autonomous helpers in human environments, recent development go... more With robots becoming increasingly autonomous helpers in human environments, recent development goes from semi-autonomous robotic systems, that need to be directed, towards autonomous partners that can cooperate with humans in joint activities. When operating autonomously in the real world, the probability of unexpected events dramatically increases and failures might make it necessary for the robot to adapt its behavior to be able to fulfill its goals. In our complex and constantly changing real world, it is not possible for the programmer to anticipate all situations a robot might encounter. So the diagnosis of cooperative plans for human robot interaction is a particular challenge since for a robot it is often unclear what is to be considered as an error. A general understanding of normality, based on the validity expectations would enable a robot to detect unexpected events and failures that have not been foreseen by the programmer, thus leading to a more robust and flexible behavior of the robot. We propose the combination of different learned models and common-sense knowledge to generate expectations, that could improve failure detection and enable us to detect and react upon unexpected events. In this paper, we formulate the key challenges for failure detection in human robot interaction, which we see in the representation of expectations, the modeling efficiency and the execution efficiency. We also provide a stack of possible knowledge-based solutions.
Workshop Proc. SOFT, 2008
Abstract. The computational tasks of model-based diagnosis and plan-ning in embedded systems can ... more Abstract. The computational tasks of model-based diagnosis and plan-ning in embedded systems can be framed as soft-constraint optimization problems with planning costs or state transition probabilities as prefer-ences. Running constraint optimization in embedded systems ...
Workshop Proc. SOFT, 2008
Abstract. The computational tasks of model-based diagnosis and plan-ning in embedded systems can ... more Abstract. The computational tasks of model-based diagnosis and plan-ning in embedded systems can be framed as soft-constraint optimization problems with planning costs or state transition probabilities as prefer-ences. Running constraint optimization in embedded systems ...
From November 7 to 12, 2010, the Dagstuhl Seminar 10451 'Runtime Verification, Diagnosis, Pla... more From November 7 to 12, 2010, the Dagstuhl Seminar 10451 'Runtime Verification, Diagnosis, Planning and Control for Autonomous Systems' was held in Schloss Dagstuhl – Leibniz Center for Informatics. During the seminar, 35 participants presented their current research and discussed ongoing work and open problems. This document puts together abstracts of the presentations given during the seminar, and provides links to extended abstracts or full papers, if available.
Annual Conference of the PHM Society, 2010
The goal of testing is to discriminate between multiple hypotheses about a system-for example, di... more The goal of testing is to discriminate between multiple hypotheses about a system-for example, different fault diagnoses of an HVAC system-by applying input patterns and verifying or falsifying the hypotheses from the observed outputs. Definitely discriminating tests (DDTs) are those input patterns that are guaranteed to discriminate between different hypotheses of nondeterministic systems. Finding DDTs is important in practice, but can be very expensive (p 2complete). Even more challenging is the problem of finding a DDT that minimizes the energy consumption of the testing process, i.e., an input pattern that can be enforced with minimal energy consumption and that is a DDT. This paper addresses both problems. We show how we can transform a given problem into a Boolean structure in decomposable negation normal form (DNNF), and extract from it a Boolean formula whose models correspond to DDTs. This allows us to harness recent advances in both knowledge compilation and satisfiability for efficient and scalable DDT computation in practice. Furthermore, we show how we can generate a DNNF structure compactly encoding all DDTs of the problem and use it to obtain an energy-optimal DDT in time linear in the size of the structure.
Vehicle-to-Grid (V2G) is the concept of buffering energy in the batteries of electric vehicles an... more Vehicle-to-Grid (V2G) is the concept of buffering energy in the batteries of electric vehicles and feeding it back into the power grid at peak demand times. The goal of this work is to as-sess the potential economic profit of V2G on different energy markets in Germany, using real market data from 2009. To this end, different energy markets in Germany are analyzed and, using a static cost model, a novel Microsoft Excel-based software tool named V2G Profit Agent was developed that allows a flexible evaluation of different V2G scenarios.
IFAC Proceedings Volumes, 1996
The paper presents objectives and results of a case study in computer support for failure mode an... more The paper presents objectives and results of a case study in computer support for failure mode and effects analysis and for the creation of repair manuals in the domain of automotive systems. Model-based prediction and diagnosis reflect the requirements of these tasks. More specifically, qualitative models of system components are necessary for both capturing the available knowledge and achieving the desired coverage and granularity of the analysis results. We describe models for parts of the anti-lock braking system (ABS) and the electronic diesel control (EDC), focusing on a qualitative approach to compositional modeling of the involved electrical circuits. The summarized results of the case study demonstrate the necessity and utility of qualitative models for the successful application of automated diagnosis to industrial problems.
Model-based diagnosis can be framed as optimization for constraints with preferences (soft constr... more Model-based diagnosis can be framed as optimization for constraints with preferences (soft constraints). We present a novel algorithm for solving soft constraints that generalizes branch-andbound search by reasoning about sets of assignments rather than individual assignments. Because in many practical cases, sets of assignments can be represented implicitly and compactly using symbolic techniques such as decision diagrams, the setbased algorithm can compute bounds faster than explicitly searching over individual assignments, while memory explosion can be avoided by limiting the size of the sets. Varying the size of the sets yields a family of algorithms that includes known search and inference algorithms as special cases. Experiments indicate that the approach can lead to significant performance improvements. 2 Constraint Optimization Problems Definition 1 (Constraint Optimization Problem) A constraint optimization problem (COP) consists of a tuple
International Joint Conference on Artificial Intelligence, 2005
Constraint optimization underlies many problems in AI. We present a novel algorithm for finite do... more Constraint optimization underlies many problems in AI. We present a novel algorithm for finite domain constraint optimization that generalizes branch-and-bound search by reasoning about sets of assignments rather than individual assignments. Because in many practical cases, sets of assign- ments can be represented implicitly and com- pactly using symbolic techniques such as deci- sion diagrams, the set-based algorithm can com-
National Conference on Artificial Intelligence, 2005
Model-based diagnosis has largely operated on hard- ware systems. However, in most complex system... more Model-based diagnosis has largely operated on hard- ware systems. However, in most complex systems to- day, hardware is augmented with software functions that inuence the system's behavior. In this paper, hard- ware models are extended to include the behavior of as- sociated embedded software, resulting in more compre- hensive diagnoses. Prior work introduced probabilistic, hierarchical, constraint-based automata (PHCA) to al-
European Conference on Artificial Intelligence, 2004
Constraint optimization is at the core of many problems in Ar- tificial Intelligence. In this pap... more Constraint optimization is at the core of many problems in Ar- tificial Intelligence. In this paper, we frame model-based diagnosis as a constraint optimization problem over lattices. We then show how it can be captured in a framework for "soft" constraints known as semiring-CSPs. The well-defined mathematical properties of a semiring-CSP permit us to devise efficient solution methods based on
Lecture Notes in Computer Science
Abstract. Many tasks in artificial intelligence, such as diagnosis, plan-ning, and reconfiguratio... more Abstract. Many tasks in artificial intelligence, such as diagnosis, plan-ning, and reconfiguration, can be framed as constraint optimization prob-lems. However, running constraint optimization within embedded sys-tems requires methods to curb the resource ...
Lecture Notes in Computer Science
Testing is the process of stimulating a system with inputs in order to reveal hidden parts of the... more Testing is the process of stimulating a system with inputs in order to reveal hidden parts of the system state. We consider a variant of the testing problem that was put forward in the model-based diagnosis literature, and consists of finding input patterns that definitely discriminate between different constraint-based system models. We show that this problem can be framed as a game played between two opponents, and naturally lends itself towards a formulation in terms of quantified CSPs. This QCSP-based formulation is a starting point to extend testing to a new classes of practically relevant applications-namely, systems with limited controllability-where tests consist of stimulation strategies instead of simple input patterns.
Lecture Notes in Computer Science, 2009
Today's complex production systems allow to simultaneously build different products following ind... more Today's complex production systems allow to simultaneously build different products following individual production plans. Such plans may fail due to component faults or unforeseen behavior, resulting in flawed products. In this paper, we propose a method to integrate diagnosis with plan assessment to prevent plan failure, and to gain diagnostic information when needed. In our setting, plans are generated from a planner before being executed on the system. If the underlying system drifts due to component faults or unforeseen behavior, plans that are ready for execution or already being executed are uncertain to succeed or fail. Therefore, our approach tracks plan execution using probabilistic hierarchical constraint automata (PHCA) models of the system. This allows to explain past system behavior, such as observed discrepancies, while at the same time it can be used to predict a plan's remaining chance of success or failure. We propose a formulation of this combined diagnosis/assessment problem as a constraint optimization problem, and present a fast solution algorithm that estimates success or failure probabilities by considering only a limited number k of system trajectories.
Uploads
Papers by Martin Sachenbacher