Golog programs allow to model complex behaviour of agents by combining primitive actions defined in a Situation Calculus theory using imperative and non-deterministic programming language constructs. In general, verifying temporal properties of Golog programs is undecidable. One way to establish decidability is to restrict the logic used by the program to a Description Logic (DL), for which recently some complexity upper bounds for verification problem have been established. However, so far it was open whether these results are tight, and lightweight DLs such as EL have not been studied at all. Furthermore, these results only apply to a setting where actions do not consume time, and the properties to be verified only refer to the timeline in a qualitative way. In a lot of applications, this is an unrealistic assumption. In this work, we study the verification problem for timed Golog programs, in which actions can be assigned differing durations, and temporal properties are specified in a metric branching time logic. This allows to annotate temporal properties with time intervals over which they are evaluated, to specify for example that some property should hold for at least n time units, or should become specified within some specified time window. We establish tight complexity bounds of the verification problem for both expressive and lightweight DLs. Our lower bounds already apply to a very limited fragment of the verification problem, and close open complexity bounds for the non-metrical cases studied before.

Automated planning is a branch of AI that addresses the problem of generating a course of action to achieve a specified objective, given an initial state of the world. It is an area that is central to the development of intelligent agents and autonomous robots. In the last decade, automated planning has seen significant progress in terms of scalability, much of it achieved by the development of heuristic search approaches. Many of these advances, are only immediately applicable to so-called classical planning tasks. However, there are compelling applications of planning that are non-classical. An example is the problem of web service composition, in which the objective is to automatically compose web artifacts to achieve the objective of a human user. In doing so, not only the hard goals but also the \emph{preferences} of the user---which are usually not considered in the classical model---must be considered. % Also, the automated composition %should deal with abstract representations of the web %artifacts---which may also not adjust to the classical model. In this thesis we show that many of the most successful advances in classical planning can be leveraged for solving compelling non-classical problems. In particular, we focus on the following non-classical planning problems: planning with temporally extended goals; planning with rich, temporally extended preferences; planning with procedural control, and planning with procedural programs that can sense the environment. We show that to efficiently solve these problems we can use a common approach: reformulation. For each of these planning tasks, we propose a reformulation algorithm that generates another, arguably simpler instance. Then, if necessary, we adapt existing techniques to make the reformulated instance solvable efficiently. In particular, we show that both the problems of planning with temporally extended goals and with procedural control can be mapped into classical planning. Planning with rich user preferences, even after reformulation, cannot be mapped into classical planning and thus we develop specialized heuristics, based on existing heuristics, together with a branch-and-bound algorithm. Finally, for the problem of planning with programs that sense, we show that under certain conditions programs can be reduced to simple operators, enabling the use of a variety of existing planners. In all cases, we show experimentally that the reformulated problems can be solved effectively by either existing planners or by our adapted planners, outperforming previous approaches.

Automated planning is a branch of AI that addresses the problem of generating a course of action to achieve a specified objective, given an initial state of the world. It is an area that is central to the development of intelligent agents and autonomous robots. In the last decade, automated planning has seen significant progress in terms of scalability, much of it achieved by the development of heuristic search approaches. Many of these advances, are only immediately applicable to so-called classical planning tasks. However, there are compelling applications of planning that are non-classical. An example is the problem of web service composition, in which the objective is to automatically compose web artifacts to achieve the objective of a human user. In doing so, not only the hard goals but also the \emph{preferences} of the user---which are usually not considered in the classical model---must be considered. % Also, the automated composition %should deal with abstract representations of the web %artifacts---which may also not adjust to the classical model. In this thesis we show that many of the most successful advances in classical planning can be leveraged for solving compelling non-classical problems. In particular, we focus on the following non-classical planning problems: planning with temporally extended goals; planning with rich, temporally extended preferences; planning with procedural control, and planning with procedural programs that can sense the environment. We show that to efficiently solve these problems we can use a common approach: reformulation. For each of these planning tasks, we propose a reformulation algorithm that generates another, arguably simpler instance. Then, if necessary, we adapt existing techniques to make the reformulated instance solvable efficiently. In particular, we show that both the problems of planning with temporally extended goals and with procedural control can be mapped into classical planning. Planning with rich user preferences, even after reformulation, cannot be mapped into classical planning and thus we develop specialized heuristics, based on existing heuristics, together with a branch-and-bound algorithm. Finally, for the problem of planning with programs that sense, we show that under certain conditions programs can be reduced to simple operators, enabling the use of a variety of existing planners. In all cases, we show experimentally that the reformulated problems can be solved effectively by either existing planners or by our adapted planners, outperforming previous approaches.

Description Logic actions specify adaptations of description logic interpretations based on some preconditions defined using a description logic. We consider DL actions in which preconditions can be specified using DL axioms as well as using conjunctive queries, and combinatiosn thereof. We investigate complexity bounds for the executability and the projection problem for these actions, which respectively ask whether an action can be executed on models of an interpretation, and which entailments are satisfied after an action has been executed on this model. In addition, we consider a set of new reasoning tasks concerned with conflicts and interactions that may arise if two action are executed at the same time. Since these problems have not been investigated before for Description Logic actions, we investigate the complexity of these tasks both for actions with conjunctive queries and without those. Finally, we consider the verification problem for Golog programs formulated over our famility of actions. Our complexity analysis considers several expressive DLs, and we provide tight complexity bounds for those for which the exact complexity of conjunctive query entailment is known.

High-level action programming languages such as Golog have successfully been used to model the behavior of autonomous agents. In addition to a logic-based action formalism for describing the environment and the effects of basic actions, they enable the construction of complex actions using typical programming language constructs. To ensure that the execution of such complex actions leads to the desired behavior of the agent, one needs to specify the required properties in a formal way, and then verify that these requirements are met by any execution of the program. Due to the expressiveness of the action formalism underlying Golog (situation calculus), the verification problem for Golog programs is in general undecidable. Action formalisms based on Description Logic (DL) try to achieve decidability of inference problems such as the projection problem by restricting the expressiveness of the underlying base logic. However, until now these formalisms have not been used within Golog programs. In the present paper, we introduce a variant of Golog where basic actions are defined using such a DL-based formalism, and show that the verification problem for such programs is decidable. This improves on our previous work on verifying properties of infinite sequences of DL actions in that it considers (finite and infinite) sequences of DL actions that correspond to (terminating and non-terminating) runs of a Golog program rather than just infinite sequences accepted by a Büchi automaton abstracting the program.

Golog is a high-level action programming language for controlling autonomous agents such as mobile robots. It is defined on top of a logic-based action theory expressed in the Situation Calculus. Before a program is deployed onto an actual robot and executed in the physical world, it is desirable, if not crucial, to verify that it meets certain requirements (typically expressed through temporal formulas) and thus indeed exhibits the desired behaviour. However, due to the high (first-order) expressiveness of the language, the corresponding verification problem is in general undecidable. In this paper, we extend earlier results to identify a large, non-trivial fragment of the formalism where verification is decidable. In particular, we consider properties expressed in a first-order variant of the branching-time temporal logic CTL*. Decidability is obtained by (1) resorting to the decidable first-order fragment C² as underlying base logic, (2) using a fragment of Golog with ground actions only, and (3) requiring the action theory to only admit local effects. / In this extended version we extend the decidability result for the verification problem to the temporal logic CTL* over C2-axioms.

The meet up between data, processes and structural knowledge in modeling complex enterprise systems is a challenging task that has led to the study of combining formalisms from knowledge representation, database theory, and process management. Moreover, to ensure system correctness, formal verification also comes into play as a promising approach that offers well-established techniques. In line with this, significant results have been obtained within the research on data-aware business processes, which studies the marriage between static and dynamic aspects of a system within a unified framework. However, several limitations are still present. Various formalisms for data-aware processes that have been studied typically use a simple mechanism for specifying the system dynamics. The majority of works also assume a rather simple treatment of inconsistency (i.e., reject inconsistent system states). Many researches in this area that consider structural domain knowledge typically also assume that such knowledge remains fixed along the system evolution (context-independent), and this might be too restrictive. Moreover, the information model of data-aware processes sometimes relies on relatively simple structures. This situation might cause an abstraction gap between the high-level conceptual view that business stakeholders have, and the low-level representation of information. When it comes to verification, taking into account all of the aspects above makes the problem more challenging. In this thesis, we investigate the verification of data-aware processes in the presence of ontologies while at the same time addressing all limitations above. Specifically, we provide the following contributions: (1) We propose a formal framework called Golog-KABs (GKABs), by leveraging on the state of the art formalisms for data-aware processes equipped with ontologies. GKABs enable us to specify semantically-rich data-aware business processes, where the system dynamics are specified using a high-level action language inspired by the Golog programming language. (2) We propose a parametric execution semantics for GKABs that is able to elegantly accommodate a plethora of inconsistency-aware semantics based on the well-known notion of repair, and this leads us to consider several variants of inconsistency-aware GKABs. (3) We enhance GKABs towards context-sensitive GKABs that take into account the contextual information during the system evolution. (4) We marry these two settings and introduce inconsistency-aware context-sensitive GKABs. (5) We introduce the so-called Alternating-GKABs that allow for a more fine-grained analysis over the evolution of inconsistency-aware context-sensitive systems. (6) In addition to GKABs, we introduce a novel framework called Semantically-Enhanced Data-Aware Processes (SEDAPs) that, by utilizing ontologies, enable us to have a high-level conceptual view over the evolution of the underlying system. We provide not only theoretical results, but have also implemented this concept of SEDAPs. We also provide numerous reductions for the verification of sophisticated first-order temporal properties over all of the settings above, and show that verification can be addressed using existing techniques developed for Data-Centric Dynamic Systems (which is a well-established data-aware processes framework), under suitable boundedness assumptions for the number of objects freshly introduced in the system while it evolves. Notably, all proposed GKAB extensions have no negative impact on computational complexity.

The meet up between data, processes and structural knowledge in modeling complex enterprise systems is a challenging task that has led to the study of combining formalisms from knowledge representation, database theory, and process management. Moreover, to ensure system correctness, formal verification also comes into play as a promising approach that offers well-established techniques. In line with this, significant results have been obtained within the research on data-aware business processes, which studies the marriage between static and dynamic aspects of a system within a unified framework. However, several limitations are still present. Various formalisms for data-aware processes that have been studied typically use a simple mechanism for specifying the system dynamics. The majority of works also assume a rather simple treatment of inconsistency (i.e., reject inconsistent system states). Many researches in this area that consider structural domain knowledge typically also assume that such knowledge remains fixed along the system evolution (context-independent), and this might be too restrictive. Moreover, the information model of data-aware processes sometimes relies on relatively simple structures. This situation might cause an abstraction gap between the high-level conceptual view that business stakeholders have, and the low-level representation of information. When it comes to verification, taking into account all of the aspects above makes the problem more challenging. In this thesis, we investigate the verification of data-aware processes in the presence of ontologies while at the same time addressing all limitations above. Specifically, we provide the following contributions: (1) We propose a formal framework called Golog-KABs (GKABs), by leveraging on the state of the art formalisms for data-aware processes equipped with ontologies. GKABs enable us to specify semantically-rich data-aware business processes, where the system dynamics are specified using a high-level action language inspired by the Golog programming language. (2) We propose a parametric execution semantics for GKABs that is able to elegantly accommodate a plethora of inconsistency-aware semantics based on the well-known notion of repair, and this leads us to consider several variants of inconsistency-aware GKABs. (3) We enhance GKABs towards context-sensitive GKABs that take into account the contextual information during the system evolution. (4) We marry these two settings and introduce inconsistency-aware context-sensitive GKABs. (5) We introduce the so-called Alternating-GKABs that allow for a more fine-grained analysis over the evolution of inconsistency-aware context-sensitive systems. (6) In addition to GKABs, we introduce a novel framework called Semantically-Enhanced Data-Aware Processes (SEDAPs) that, by utilizing ontologies, enable us to have a high-level conceptual view over the evolution of the underlying system. We provide not only theoretical results, but have also implemented this concept of SEDAPs. We also provide numerous reductions for the verification of sophisticated first-order temporal properties over all of the settings above, and show that verification can be addressed using existing techniques developed for Data-Centric Dynamic Systems (which is a well-established data-aware processes framework), under suitable boundedness assumptions for the number of objects freshly introduced in the system while it evolves. Notably, all proposed GKAB extensions have no negative impact on computational complexity.

The Golog action programming language is a powerful means to express high-level behaviours in terms of programs over actions defined in a Situation Calculus theory. In particular for physical systems, verifying that the program satisfies certain desired temporal properties is often crucial, but undecidable in general, the latter being due to the language’s high expressiveness in terms of first-order quantification and program constructs. So far, approaches to achieve decidability involved restrictions where action effects either had to be contextfree (i.e. not depend on the current state), local (i.e. only affect objects mentioned in the action’s parameters), or at least bounded (i.e. only affect a finite number of objects). In this paper, we present a new, more general class of action theories (called acyclic) that allows for context-sensitive, non-local, unbounded effects, i.e. actions that may affect an unbounded number of possibly unnamed objects in a state-dependent fashion. We contribute to the further exploration of the boundary between decidability and undecidability for Golog, showing that for acyclic theories in the two-variable fragment of first-order logic, verification of CTL properties of programs over ground actions is decidable.

This dissertation investigates high-level decision making for agents that are both goal and utility driven. We develop a partially observable Markov decision process (POMDP) planner which is an extension of an agent programming language called DTGolog, itself an extension of the Golog language. Golog is based on a logic for reasoning about action—the situation calculus. A POMDP planner on its own cannot cope well with dynamically changing environments and complicated goals. This is exactly a strength of the belief-desire-intention (BDI) model: BDI theory has been developed to design agents that can select goals intelligently, dynamically abandon and adopt new goals, and yet commit to intentions for achieving goals. The contribution of this research is twofold: (1) developing a relational POMDP planner for cognitive robotics, (2) specifying a preliminary BDI architecture that can deal with stochasticity in action and perception, by employing the planner. / Computing / M. Sc. (Computer Science)