Bernhard K. Aichernig

Martin Tappler, Filip Cano Córdoba, Bernhard K. Aichernig et al.

Evaluation of deep reinforcement learning (RL) is inherently challenging. Especially the opaqueness of learned policies and the stochastic nature of both agents and environments make testing the behavior of deep RL agents difficult. We present a search-based testing framework that enables a wide range of novel analysis capabilities for evaluating the safety and performance of deep RL agents. For safety testing, our framework utilizes a search algorithm that searches for a reference trace that solves the RL task. The backtracking states of the search, called boundary states, pose safety-critical situations. We create safety test-suites that evaluate how well the RL agent escapes safety-critical situations near these boundary states. For robust performance testing, we create a diverse set of traces via fuzz testing. These fuzz traces are used to bring the agent into a wide variety of potentially unknown states from which the average performance of the agent is compared to the average performance of the fuzz traces. We apply our search-based testing approach on RL for Nintendo's Super Mario Bros.

Edi Muskardin, Martin Tappler, Bernhard K. Aichernig et al.

In practical applications, we can rarely assume full observability of a system's environment, despite such knowledge being important for determining a reactive control system's precise interaction with its environment. Therefore, we propose an approach for reinforcement learning (RL) in partially observable environments. While assuming that the environment behaves like a partially observable Markov decision process with known discrete actions, we assume no knowledge about its structure or transition probabilities. Our approach combines Q-learning with IoAlergia, a method for learning Markov decision processes (MDP). By learning MDP models of the environment from episodes of the RL agent, we enable RL in partially observable domains without explicit, additional memory to track previous interactions for dealing with ambiguities stemming from partial observability. We instead provide RL with additional observations in the form of abstract environment states by simulating new experiences on learned environment models to track the explored states. In our evaluation, we report on the validity of our approach and its promising performance in comparison to six state-of-the-art deep RL techniques with recurrent neural networks and fixed memory.

Edi Muškardin, Martin Tappler, Ingo Pill et al.

We examine the assumption that the hidden-state vectors of recurrent neural networks (RNNs) tend to form clusters of semantically similar vectors, which we dub the clustering hypothesis. While this hypothesis has been assumed in the analysis of RNNs in recent years, its validity has not been studied thoroughly on modern neural network architectures. We examine the clustering hypothesis in the context of RNNs that were trained to recognize regular languages. This enables us to draw on perfect ground-truth automata in our evaluation, against which we can compare the RNN's accuracy and the distribution of the hidden-state vectors. We start with examining the (piecewise linear) separability of an RNN's hidden-state vectors into semantically different classes. We continue the analysis by computing clusters over the hidden-state vector space with multiple state-of-the-art unsupervised clustering approaches. We formally analyze the accuracy of computed clustering functions and the validity of the clustering hypothesis by determining whether clusters group semantically similar vectors to the same state in the ground-truth model. Our evaluation supports the validity of the clustering hypothesis in the majority of examined cases. We observed that the hidden-state vectors of well-trained RNNs are separable, and that the unsupervised clustering techniques succeed in finding clusters of similar state vectors.

Paul Kobialka, Andrea Pferscher, Bernhard K. Aichernig et al.

With the servitization of business, understanding how users experience services becomes a crucial success factor for companies. Therefore, there is a need to include feedback from user experiences in the software engineering process. Behavioral models of user journeys, describing how users experience their interaction with a service, can provide insights and potentially improve services. In this paper, we investigate techniques that allow the automatic generation of behavioral models from user interactions with a service, recorded in an event log. We first compare two established techniques that generate behavioral models from a given event log: automata learning and process mining. Afterward, we present a novel, hybrid method that combines both automata learning and process mining methods to overcome their limitations. For the existing techniques, we present methods to learn models of user journeys and evaluate the accuracy of the resulting models. We then compare these techniques with our novel method for the automatic extraction of user journey models from the event logs of digital services. We assess the practical applicability of all techniques by evaluating real-world applications. Our results show that process mining techniques rely on expert knowledge, while automata learning techniques depend on the distribution of events in the given event log. We further show that the proposed hybrid technique combines the strengths of both process mining and automata learning, automatically selecting the best method and parameter settings for a given event log to learn very accurate models.

Martin Tappler, Edi Muškardin, Bernhard K. Aichernig et al.

Solving control tasks in complex environments automatically through learning offers great potential. While contemporary techniques from deep reinforcement learning (DRL) provide effective solutions, their decision-making is not transparent. We aim to provide insights into the decisions faced by the agent by learning an automaton model of environmental behavior under the control of an agent. However, for most control problems, automata learning is not scalable enough to learn a useful model. In this work, we raise the capabilities of automata learning such that it is possible to learn models for environments that have complex and continuous dynamics. The core of the scalability of our method lies in the computation of an abstract state-space representation, by applying dimensionality reduction and clustering on the observed environmental state space. The stochastic transitions are learned via passive automata learning from observed interactions of the agent and the environment. In an iterative model-based RL process, we sample additional trajectories to learn an accurate environment model in the form of a discrete-state Markov decision process (MDP). We apply our automata learning framework on popular RL benchmarking environments in the OpenAI Gym, including LunarLander, CartPole, Mountain Car, and Acrobot. Our results show that the learned models are so precise that they enable the computation of policies solving the respective control tasks. Yet the models are more concise and more general than neural-network-based policies and by using MDPs we benefit from a wealth of tools available for analyzing them. When solving the task of LunarLander, the learned model even achieved similar or higher rewards than deep RL policies learned with stable-baselines3.

Benjamin von Berg, Bernhard K. Aichernig

AALpy is a well-established open-source automata learning library written in Python with a focus on active learning of systems with IO behavior. It provides a wide range of state-of-the-art algorithms for different automaton types ranging from fully deterministic to probabilistic automata. In this work, we present the recent addition of a generalized implementation of an important method from the domain of passive automata learning: state-merging in the red-blue framework. Using a common internal representation for different automaton types allows for a general and highly configurable implementation of the red-blue framework. We describe how to define and execute state-merging algorithms using AALpy, which reduces the implementation effort for state-merging algorithms mainly to the definition of compatibility criteria and scoring. This aids the implementation of both existing and novel algorithms. In particular, defining some existing state-merging algorithms from the literature with AALpy only takes a few lines of code.

Bernhard K. Aichernig, Roderick Bloem, Masoud Ebrahimi et al.

Models play an essential role in the design process of cyber-physical systems. They form the basis for simulation and analysis and help in identifying design problems as early as possible. However, the construction of models that comprise physical and digital behavior is challenging. Therefore, there is considerable interest in learning such hybrid behavior by means of machine learning which requires sufficient and representative training data covering the behavior of the physical system adequately. In this work, we exploit a combination of automata learning and model-based testing to generate sufficient training data fully automatically. Experimental results on a platooning scenario show that recurrent neural networks learned with this data achieved significantly better results compared to models learned from randomly generated data. In particular, the classification error for crash detection is reduced by a factor of five and a similar F1-score is obtained with up to three orders of magnitude fewer training samples.

Martin Tappler, Bernhard K. Aichernig, Giovanni Bacci et al.

Automata learning techniques automatically generate system models from test observations. These techniques usually fall into two categories: passive and active. Passive learning uses a predetermined data set, e.g., system logs. In contrast, active learning actively queries the system under learning, which is considered more efficient. An influential active learning technique is Angluin's L* algorithm for regular languages which inspired several generalisations from DFAs to other automata-based modelling formalisms. In this work, we study L*-based learning of deterministic Markov decision processes, first assuming an ideal setting with perfect information. Then, we relax this assumption and present a novel learning algorithm that collects information by sampling system traces via testing. Experiments with the implementation of our sampling-based algorithm suggest that it achieves better accuracy than state-of-the-art passive learning techniques with the same amount of test data. Unlike existing learning algorithms with predefined states, our algorithm learns the complete model structure including the states.

Martin Tappler, Bernhard K. Aichernig, Roderick Bloem

This paper presents a learning-based approach to detecting failures in reactive systems. The technique is based on inferring models of multiple implementations of a common specification which are pair-wise cross-checked for equivalence. Any counterexample to equivalence is flagged as suspicious and has to be analysed manually. Hence, it is possible to find possible failures in a semi-automatic way without prior modelling. We show that the approach is effective by means of a case study. For this case study, we carried out experiments in which we learned models of five implementations of MQTT brokers/servers, a protocol used in the Internet of Things. Examining these models, we found several violations of the MQTT specification. All but one of the considered implementations showed faulty behaviour. In the analysis, we discuss effectiveness and also issues we faced.

Martin Tappler, Bernhard K. Aichernig, Kim Guldstrand Larsen et al.

Model learning has gained increasing interest in recent years. It derives behavioural models from test data of black-box systems. The main advantage offered by such techniques is that they enable model-based analysis without access to the internals of a system. Applications range from fully automated testing over model checking to system understanding. Current work focuses on learning variations of finite state machines. However, most techniques consider discrete time. In this paper, we present a method for learning timed automata, finite state machines extended with real-valued clocks. The learning method generates a model consistent with a set of timed traces collected by testing. This generation is based on genetic programming, a search-based technique for automatic program creation. We evaluate our approach on 44 timed systems, comprising four systems from the literature and 40 randomly generated examples.

Bernhard K. Aichernig, Elisabeth Jöbstl

Model-based mutation testing uses altered test models to derive test cases that are able to reveal whether a modelled fault has been implemented. This requires conformance checking between the original and the mutated model. This paper presents an approach for symbolic conformance checking of action systems, which are well-suited to specify reactive systems. We also consider nondeterminism in our models. Hence, we do not check for equivalence, but for refinement. We encode the transition relation as well as the conformance relation as a constraint satisfaction problem and use a constraint solver in our reachability and refinement checking algorithms. Explicit conformance checking techniques often face state space explosion. First experimental evaluations show that our approach has potential to outperform explicit conformance checkers.