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.

This paper describes the New Millennium Remote Agent #NMRA# architecture for autonomous spacecraft control systems. This architecture integrates traditional real-time monitoring and control with constraintbased planning and scheduling, robust multi-threaded execution, and model-based diagnosis and recon#guration.

Planetary rovers provide a considerable challenge for robotic systems in that they must operate for long periods autonomously, or with relatively little intervention. To achieve this, they need to have on-board fault detection and diagnosis capabilities in order to determine the actual state of the vehicle, and decide what actions are safe to perform. Traditional model-based diagnosis techniques are not suitable for rovers due to the tight coupling between the vehicle’s performance and its environment. Hybrid diagnosis using particle filters is presented as an alternative, and its strengths and weaknesses are examined. We also present some extensions to particle filters that are designed to make them more suitable for use in diagnosis problems. 1

Filtering denotes any method whereby an agent updates its belief state—its knowledge of the state of the world—from a sequence of actions and observations. In logical filtering, the belief state is a logical formula describing possible world states and the agent has a (possibly nondeterministic) logical model of its environment and sensors. This paper presents efficient logical filtering algorithms that maintain a compact belief state representation indefinitely, for a broad range of environment classes including nondeterministic, partially observable STRIPS environments and environments in which actions permute the state space. Efficient filtering is also possible when the belief state is represented using prime implicates, or when it is approximated by a logically weaker formula. 1

Deep Space One will be the first spacecraft to be controlled by an autonomous agent potentially capable of carrying out a complete mission with minimal commanding from Earth. The New Millennium Remote Agent (NMRA) includes a planner-scheduler that produces plans, and an executive that carries them out. In this paper we discuss several issues arising at the interface between planning and execution. including execution latency, plan dispatchability, and the distinction between controllable and uncontrollable events. Temporal information in the plan is represented within the general framework of Simple Temporal Constraint networks, as introduced by Dechter, Meiri, and Pearl. However, the execution requirements have a substantial impact on the topology of the links and the propagation through the network. 1 Introduction Deep Space One (DS1) will be the first spacecraft to be controlled by an autonomous agent, the New Millennium Remote Agent (NMRA) [15], potentially capable of carrying ou...

Most systems for model--based reasoning record justifications during the search for solutions. In the popular ATMS systems this can lead to an explosion of recorded information when solving complicated problems. We propose an engine for model-based reasoning which works directly on logical models, uses static precompiled information on the structure of the underlying theory and does no further recording during search. This engine (DRUM-II) solves large complicated problems with attractive time and space complexity. To demonstrate the efficiency of the engine we give a new characterization of a popular benchmark suite, solve it with our engine and compare the performance to previous results.

In recent years, there has been an explosion of research in AI into propositional satis ability (or Sat). There are many factors behind the increased interest in this area. One factor is the improvement in search procedures for Sat. New local search procedures like Gsat are able to solve Sat problems with thousands of variables. At the same time, implementations of complete search algorithms like Davis-Putnam have been able to solve open mathematical problems. Another factor is the identi cation of hard Sat problems at a phase transition in solubility. A third factor is the demonstration that we can often solve real world problems by encoding them into Sat. There has also seen an improved theoretical understanding of Sat, particularly in the analysis of such phase transition behaviour. This paper reviews the state of the art for research into satis ability, and discuss applications in which algorithms for satis ability have proved successful.

Fault detection and isolation are critical tasks to ensure correct operation of systems. When we consider stochastic hybrid systems, diagnosis algorithms need to track both the discrete mode and the continuous state of the system in the presence of noise. Deterministic techniques like Livingstone cannot deal with the stochasticity in the system and models. Conversely Bayesian belief update techniques such as particle filters may require many computational resources to get a good approximation of the true belief state. In this paper we propose a fault detection and isolation architecture for stochastic hybrid systems that combines look-ahead RaoBlackwellized Particle Filters (RBPF) with the Livingstone 3 (L3) diagnosis engine. In this approach RBPF is used to track the nominal behavior, a novel n-step prediction scheme is used for fault detection and L3 is used to generate a set of candidates that are consistent with the discrepant observations which then continue to be tracked by the RBPF scheme.

The Remote Agent (RA) integrates a broad spectrum of robotic activities, including planning, scheduling, execution, monitoring, failure detection, diagnosis, and recovery. The RA Executive (EXEC) can be viewed as the core of the agent. EXEC enables software developers to think about the robot at a higher level; it also supports the reuse of knowledge and code across multiple robot applications. EXEC's capabilities include a high-level procedural action-definition language, services for resource management and configuration management, support for executing flexible closed-loop plans and command sequences, support for replanning, and a framework for specifying fault responses including safe modes and responses to plan failures. We believe that these capabilities are required for most autonomous agents, and that executives and execution capabilities will become more essential as we attempt to develop autonomous agents of increasing capability. Moreover, we deem modular...

This paper describes the validation of the Remote Agent Experiment. A primary goal of this experiment was to provide an onboard demonstration of spacecraft autonomy. This demonstration included both nominal operations with goal-oriented commanding and closed-loop plan execution, and fault protection capabilities with failure diagnosis and recovery, on-board replanning following unrecoverable failures, and system-level fault protection. Other equally important goals of the experiment were to decrease the risk of deploying Remote Agents on future missions and to familiarize the spacecraft engineering community with the Remote Agent approach. These goals were achieved by successfully integrating the Remote Agent with the Deep Space 1 flight software, developing a layered testing approach, and taking various steps to gain the confidence of the spacecraft team. In this paper we describe how we achieved our goals, and discuss the actual on-board demonstration in May, 199...