An ASP Framework for Reactive Planning and Monitoring

Planning plays an important role in many relevant areas of Artificial Intelligence. However, in real world applications the classical planning problem (i.e., finding a sequence of actions to execute in order to reach a certain goal) is often only of limited relevance as soon as exogenous events can change the state of the world or unforeseen effects can arise from taking an action during execution. On the other hand, especially in robotics and software agents, approaches with pure reactive behavior and no planning capabilities are often inflexible and allow only for executing hard-coded "plans" from a fixed library. Obviously, hybrid approaches are necessary, where planning, monitoring and reacting to the real world are interleaved. In particular, this plays a role when planning for collaborative tasks of agents. The main target of this project is to combine and extend logic based approaches on planning and monitoring in a common framework based on answer set programming, which is a growing field in declarative logic programming. Fundamental research in the areas of planning, diagnosis and belief revision has shown promising and competitive results by use of logic programming for elementary tasks and problem solving in the respective fields separately. However, all these fields are important for a reactive agent environment. Integrating respective methods from these fields is essential for successful applications of planning in real world environments. To this end, we can take advantage of the knowledge and experience that we have gained in previous projects, including the FWF projects P11580-MAT, P14781-INF and P13871-INF. We intend to use the DLV system and the planning system DLV^K based on DLV, that have been developed in the quoted projects. DLV is a state-of- the-art system for disjunctive logic programming. It is used as an efficient computational engine which, together with the planning system DLV^K, will serve as the basis for building the implementation.

Planning plays an important role in many relevant areas of Artificial Intelligence. However, in real world applications the classical planning problem (i.e., finding a sequence of actions to execute in order to reach a certain goal) is often only of limited relevance as soon as exogenous events can change the state of the world or unforeseen effects can arise from taking an action during execution. On the other hand, especially in robotics and software agents, approaches with pure reactive behavior and no planning capabilities are often inflexible and allow only for executing hard-coded "plans" from a fixed library. Obviously, hybrid approaches are necessary, where planning, monitoring and reacting to the real world are interleaved. In particular, this plays a role when planning for collaborative tasks of agents. The main target of this project is to combine and extend logic based approaches on planning and monitoring in a common framework based on answer set programming, which is a growing field in declarative logic programming. Fundamental research in the areas of planning, diagnosis and belief revision has shown promising and competitive results by use of logic programming for elementary tasks and problem solving in the respective fields separately. However, all these fields are important for a reactive agent environment. Integrating respective methods from these fields is essential for successful applications of planning in real world environments. To this end, we can take advantage of the knowledge and experience that we have gained in previous projects, including the FWF projects P11580-MAT, P14781-INF and P13871-INF. We intend to use the DLV system and the planning system DLV^K based on DLV, that have been developed in the quoted projects. DLV is a state-of- the-art system for disjunctive logic programming. It is used as an efficient computational engine which, together with the planning system DLV^K, will serve as the basis for building the implementation.