Heesch, René

Industrial fault diagnosis exhibits the perennial problem of reasoning with partial and real-valued information. This is mainly due to the fact that in real-world applications, industrial systems are only instrumented insofar, as sensor information is required for their functioning. However, such instrumentation leaves out much information that would be useful for fault diagnosis. This is problematic since consistency-based fault diagnosis uses available information and computes intermediate values within a system description. These values are then used to compare expected normal behaviour to actual observed values. In the past, this was done only for Boolean circuits. Recently, satisfiability modulo non-linear arithmetic (SMT) formulations have been developed that allow the calculation of real values, instead of only Boolean ones. Leveraging those formulations, we in this article present a novel method to infer missing sensor values using an SMT system description and the notion of critical pairs. We show on a running example and also empirically that we can infer novel measurements for five process industrial systems. We conclude that, although SMT calculations accumulate some error, we can infer novel optimal measurements for all systems.

This is an extended abstract of the manuscript "A Lazy Approach to Neural Numerical Planning with Control Parameters". The paper presents a lazy, hierarchical approach to tackle the challenge of planning in complex numerical domains, where the effects of actions are influenced by control parameters, and may be described by neural networks.

Approaches in AI planning for Cyber-Physical Production Systems (CPPS) are mainly symbolic and depend on comprehensive formalizations of system domains and planning problems. Handcrafting such formalizations requires detailed knowledge of the formalization language, of the CPPS, and is overall considered difficult, tedious, and error-prone. Within this paper, we suggest a sub-symbolic approach for solving planning problems in CPPS. Our approach relies on neural networks that learn the dynamical behavior of individual process steps from global time-series observations of the CPPS and are embedded in a superordinate network architecture. In this context, we present the process step representation network architecture (peppr), a novel neural network architecture, which can learn the behavior of individual or multiple dynamical systems from global time-series observations. We evaluate peppr on real datasets from physical and biochemical CPPS, as well as artificial datasets from electrical and mathematical domains. Our model outperforms baseline models like multilayer perceptrons and variational autoencoders and can be considered as a first step towards a sub-symbolic approach for planning in CPPS.

In this paper, we tackle the problem of planning in complex numerical domains, where actions are indexed by control parameters, and their effects may be described by neural networks. We propose a lazy, hierarchical approach based on two ingredients. First, a Satisfiability Modulo Theory solver looks for an abstract plan where the neural networks in the model are abstracted into uninterpreted functions. Then, we attempt to concretize the abstract plan by querying the neural network to determine the control parameters. If the concretization fails and no valid control parameters could be found, suitable information to refine the abstraction is lifted to the Satisfiability Modulo Theory model. We contrast our work against the state of the art in NN-enriched numerical planning, where the neural network is eagerly and exactly represented as terms in Satisfiability Modulo Theories over nonlinear real arithmetic. Our systematic evaluation on four different planning domains shows that avoiding symbolic reasoning about the neural network not only leads to substantial efficiency improvements, but also enables their integration as black-box models.

Automated planning approaches provide robust and efficient methods to automatically find plans for a given problem and a set of possible actions. However, due to the rather high effort required to create planning models, these approaches cannot be used for adaptable manufacturing plants. In this contribution, we present a method to automatically generate a planning problem in the form of PDDL from an existing capability model. This method eliminates the additional effort required to model a planning problem, making planning approaches usable for adaptable manufacturing plants.