Renewed motives for space exploration have inspired NASA to work toward the goal of establishing a virtual presence in space, through heterogeneous effets of robotic explorers. Information technology, and Artificial Intelligence in particular, will play a central role in this endeavor by endowing these explorers with a form of computational intelligence that we call remote agents. In this paper we describe the Remote Agent, a specific autonomous agent architecture based on the principles of model-based programming, on-board deduction and search, and goal-directed closed-loop commanding, that takes a significant step toward enabling this future. This architecture addresses the unique characteristics of the spacecraft domain that require highly reliable autonomous operations over long periods of time with tight deadlines, resource constraints, and concurrent activity among tightly coupled subsystems. The Remote Agent integrates constraint-based temporal planning and scheduling, robust multi-threaded execution, and model-based mode identification and reconfiguration. The demonstration of the integrated system as an on-board controller for Deep Space One, NASA's rst New Millennium mission, is scheduled for a period of a week in late 1998. The development of the Remote Agent also provided the opportunity to reassess some of AI's conventional wisdom about the challenges of implementing embedded systems, tractable reasoning, and knowledge representation. We discuss these issues, and our often contrary experiences, throughout the paper.

Artificial intelligence has traditionally used constraint satisfaction and logic to frame a wide range of problems, including planning, diagnosis, cognitive robotics and embedded systems control. However, many decision making problems are now being re-framed as optimization problems, involving a search over a discrete space for the best solution that satisfies a set of constraints. The best methods for finding optimal solutions, such as A*, explore the space of solutions one state at a time. This paper introduces conflict-directed A*, a method for solving optimal constraint satisfaction problems. Conflict-directed A* searches the state space in best first order, but accelerates the search process by eliminating subspaces around each state that are inconsistent. This elimination process builds upon the concepts of conflict and kernel diagnosis used in model-based diagnosis[1,2] and in dependency-directed search[3--6]. Conflict-directed A* is a fundamental tool for building model-based embedded systems, and has been used to solve a range of problems, including fault isolation[1], diagnosis[7], mode estimation and repair[8], model-compilation[9] and model-based programming[10].

. A problem-solving method describes a reasoning process that efficiently achieves a goal by applying domain knowledge. However, a problem-solving method cannot directly be applied because of the existence of a gap between, on the one hand, a problem-solving method and the domain knowledge it uses, and, on the other hand, a problem-solving method and the goal that it is supposed to achieve. In this paper, we distinguish two types of assumptions based on an architecture of problem-solving methods, that are able to bridge the gap: one type of assumption is used to strengthen a problem-solving method, and the other to weaken the goal to be achieved. We also show how the effect of one assumption type can be substituted by the effect of the other type, and refer to this as "the law of conservation of assumptions". 1 Introduction The notion of problem-solving method (PSM) is present in many current knowledge engineering frameworks such as Generic Tasks [6], Role-Limiting Methods [15], KADS ...

We present a paradigmatic example of a feedback-controlled system: an electric motor with sensor and controller. Diagnosis of this system is performed based on a qualitative model that reflects deviations of parameters and behavior from a fixed reference state. The hypothesis

In this paper, we use a logic-based system description for a simple (non-logic) functional language to examine the ways in which a diagnosis system can use its system description to improve debugging performance. The key concept is that the notion of expression replacement, which is the basis for repairing a program, can also serve as a fundamental heuristic for searching the source of an error. We formally define replacements in terms of fault modes, explicitly define a replacement order, and use the replacement heuristic for finding diagnoses. Finally, we incorporate the use of multiple test cases and discuss their use in discriminating between diagnoses. 1 Introduction Although model-based diagnosis (MBD) [Reiter, 1987; de Kleer and Williams, 1987] provides a general diagnosis framework, most of its applications are within the hardware domain or use models of the physical world. However, several attempts have been made to extend the application of diagnosis methods to software debu...

Diagnosis was among the first subjects investigated when digital computers became available. It still remains an important research area, in which several new developments have taken place in the last decade. One of these new developments is the use of detailed domain models in knowledge-based systems for the purpose of diagnosis, often referred to as model-based diagnosis. Typically, such models embody knowledge of the normal or abnormal structure and behaviour of the modelled objects in a domain. Models of the structure and workings of technical devices, and causal models of disease processes in medicine are two examples. In this article, the most important notions of diagnosis and their formalisation are reviewed and brought in perspective. In addition, attention is focused on a number of general frameworks of diagnosis, which offer sufficient flexibility for expressing several types of diagnosis.

. This paper presents a set of focusing methods for model-based diagnosis systems. They aim at restricting the efforts spent on different computational tasks: the generation of diagnosis candidates, model-based prediction, and dependency recording. Building upon our previous work on exploiting a preference order on component faults in the default-based diagnosis engine (DDE), we formally describe the focusing principles and show their validity in a default logic framework. 1 INTRODUCTION Understanding how complex devices work is difficult enough. Diagnosing them when they don't, is still much harder, because many different faulty behaviors can be exhibited by each system constituent and, in a combinatorial way, by the entire system. Nevertheless, human experts often manage to navigate through this huge space of possibilities quite economically. This economy is essentially grounded on a focusing principle: "Do (or consider) only what appears necessary for the case at hand". Applied to...

Preferences and strategies are fundamental to model-based diagnosis, for specifying preferred and fall-back approaches to the diagnosis task, both to capture general and domain specific criteria, but also to tackle the complexity issue by employing heuristics. A formal framework based on extended logic programming and meta-programs is provided to represent preferences and strategies required by model-based diagnosis. This framework is clearer and more expressive than other approaches that have addressed these problems. We show how the concepts of preferences and strategies are directly programmed and captured by logic meta-programming and meta-reasoning methods, and their implementation techniques. The paper is intended as proof-of-principle that all concepts needed by a model-based diagnosis system can represented declaratively and captured by a logic meta-program. Specialized more efficient algorithms can be substituted for the simpler proof-of-principle ones we include, and are the...

Model-based diagnosis (MBD) tackles the problem of troubleshooting systems starting from a description of their structure and function (or behavior). Time is a fundamental dimension in MBD: the behavior of most systems is time-dependent in one way or another. Temporal MBD, however, is a difficult task and indeed many simplifying assumptions have been adopted in the various approaches in the literature. These assumptions concern different aspects such as the type and granularity of the temporal phenomena being modeled, the definition of diagnosis, the ontology for time being adopted. Unlike the atemporal case, moreover, there is no general "theory" of temporal MBD which can be used as a knowledge-level characterization of the problem. In this paper we present a general characterization of temporal model-based diagnosis. We distinguish between different temporal phenomena that can be taken into account in diagnosis and we introduce a modeling language which can capture all such phenomena...

Most current model-based diagnosis formalisms and algorithms are defined only for static systems, which is often inadequate for medical reasoning. In this paper we describe a model-based framework plus algorithms for diagnosing timedependent systems where we can define qualitative temporal scenarios. Complex temporal behavior is described within a logical framework extended by qualitative temporal constraints. Abstract observations aggregate from observations at time points to assumptions over time intervals. These concepts provide a very natural representation and make diagnosis independent of the number of actual observations and the temporal resolution. The concept of abstract temporal diagnosis captures in a natural way the kind of indefinite temporal knowledge which is frequently available in medical diagnoses. We use viral Hepatitis B (including a set of real Hepatitis B data) to illustrate and evaluate our framework. The comparison of our results with the results of HEPAXPERT-I is promising. The diagnosis computed in our system is often more precise than the diagnosis in HEPAXPERT-I and we detect inconsistent data sequences which cannot be detected in the latter system.