This paper addresses the formal verification of diagnosis systems. We tackle the problem of diagnosability: given a partially observable dynamic system, and a diagnosis system observing its evolution over time, we discuss how to verify (at design time) if the diagnosis system will be able to infer (at runtime) the required information on the hidden part of the dynamic state. We tackle the problem by looking for pairs of scenarios that are observationally indistinguishable, but lead to situations that are required to be distinguished. We reduce the problem to a model checking problem. The finite state machine modeling the dynamic system is replicated to construct such pairs of scenarios; the diagnosability conditions are formally expressed in temporal logic; the check for diagnosability is carried out by solving a model checking problem. We focus on the practical applicability of the method. We show how the formalism is adequate to represent diagnosability problems arising from a significant, real-world application. Symbolic model checking techniques are used to formally verify and incrementally refine the diagnosability conditions. 1

In a field of research about general reasoning mechanisms, it is essential to have appropriate benchmarks. Ideally, the benchmarks should reflect possible applications of the developed technology. In AI Planning, researchers more and more tend to draw their testing examples from the benchmark collections used in the International Planning Competition (IPC). In the organization of (the deterministic part of) the fourth IPC, IPC-4, the authors therefore invested significant effort to create a useful set of benchmarks. They come from five different (potential) real-world applications of planning: airport ground traffic control, oil derivative transportation in pipeline networks, model-checking safety properties, power supply restoration, and UMTS call setup. Adapting and preparing such an application for use as a benchmark in the IPC involves, at the time, inevitable (often drastic) simplifications, as well as careful choice between, and engineering of, domain encodings. For the first time in the IPC, we used compilations to formulate complex domain features in simple languages such as STRIPS, rather than just dropping the more interesting problem constraints in the simpler language subsets. The article explains and discusses the five application domains and their adaptation to form the PDDL test suites used in IPC-4. We summarize known

We present a new algorithm for probabilistic planning with no observability. Our algorithm, called Probabilistic-FF, extends the heuristic forward-search machinery of Conformant-FF to problems with probabilistic uncertainty about both the initial state and action effects. Specifically, Probabilistic-FF combines Conformant-FF’s techniques with a powerful machinery for weighted model counting in (weighted) CNFs, serving to elegantly define both the search space and the heuristic function. Our evaluation of Probabilistic-FF on several probabilistic domains shows an unprecedented, several orders of magnitude improvement over previous results in this area.

We present a case study in confronting the GPT generalpurpose planner with the challenging power supply restoration (PSR) benchmark for contingent planning. PSR is derived from a real-world problem, and the difculty of modeling and solving it contrasts with that of the purely articial benchmarks commonly used in the literature. This confrontation leads us to improve general techniques for contingent planning, to provide a PDDL-syle encoding of PSR which we hope to see used in planning competitions, and to report the rst results on generating optimal policies for PSR.

We improve the efficiency of Sampath's diagnoser approach by exploiting compact symbolic representations of the system and diagnoser in terms of binary decision diagrams. We present an algorithm for synthesising the symbolic diagnoser with promising results on test cases derived from a telecommunication application.

Abstract. Given a configuration of parameters that satisfies a set of constraints, and given external changes that change and fix the value of some parameters making the configuration invalid, the problem of interactive reconfiguration is to assist a user to interactively reassign a subset of the parameters to reach a consistent configuration again. In this paper, we present two BDD-based algorithms for solving the problem, one based on a monolithic BDD-representation of the solution space and another using a set of BDDs. We carry out experiments on a set of power supply restoration benchmarks and show that the set-of-BDDs algorithm scales much better. 1

We present a new algorithm for probabilistic planning with no observability. Our algorithm, called Probabilistic-FF, extends the heuristic forward-search machinery of Conformant-FF to problems with probabilistic uncertainty about both the initial state and action effects. Specifically, Probabilistic-FF combines Conformant-FF’s techniques with a powerful machinery for weighted model counting in (weighted) CNFs, serving to elegantly define both the search space and the heuristic function. Our evaluation of Probabilistic-FF shows its fine scalability in a range of probabilistic domains, constituting a several orders of magnitude improvement over previous results in this area. We use a problematic case to point out the main open issue to be addressed by further research. 1.

There is controversy as to whether explicit support for PDDL-like axioms and derived predicates is needed for planners to handle real-world domains effectively. Many researchers have deplored the lack of precise semantics for such axioms, while others have argued that it might be best to compile them away. We propose an adequate semantics for PDDL axioms and show that they are an essential feature by proving that it is impossible to compile them away if we restrict the growth of plans and domain descriptions to be polynomial. These results suggest that adding a reasonable implementation to handle axioms inside the planner is beneficial for the performance. Our experiments confirm this suggestion.

Distributed constraint optimization (DCOP) provides a framework for coordinated decision making by a team of agents. Often, during the decision making, capacity constraints on agents ’ resource consumption must be taken into account. To address such scenarios, an extension of DCOP- Resource Constrained DCOP- has been proposed. However, certain type of resources have an additional structure associated with them and exploiting it can result in more efficient algorithms than possible with a general framework. An example of these are distribution networks, where the flow of a commodity from sources to sinks is limited by the flow capacity of edges. We present a new model of structured resource constraints that exploits the acyclicity and the flow conservation property of distribution networks. We show how this model can be used in efficient algorithms for finding the optimal flow configuration in distribution networks, an essential problem in managing power distribution networks. Experiments demonstrate the efficiency and scalability of our approach on publicly available benchmarks and compare favorably against a specialized solver for this task. Our results extend significantly the effectiveness of distributed constraint optimization for practical multi-agent settings.

Non-determinism is often caused by infrequent errors that make otherwise deterministic actions fail. In this paper, we introduce fault tolerant planning to address this problem. An n-fault tolerant plan is guaranteed to recover from up to n errors occurring during its execution. We show how optimal n-fault tolerant plans can be generated via the strong universal planning algorithm. This algorithm uses an implicit search technique based on the reduced Ordered Binary Decision Diagram (OBDD) that is particularly well suited for non-deterministic planning and has outperformed most alternative approaches. However, the OBDDs used to represent the blind backward search of the strong algorithm often blow up. A heuristic version of the algorithm has recently been proposed but is incapable of dynamically guiding the recovery part of the plan toward error states. To address this problem, we introduce two specialized algorithms 1-FTP (blind) and 1-GFTP (guided) for 1-fault tolerant planning that decouples the synthesis of the recovery and nonrecovery part of the plan. Our experimental evaluation includes 7 domains of which 3 are significant real-world cases. It verifies