informatik-vortraege Januar 2023

informatik-vortraege@lists.rwth-aachen.de

3 Teilnehmer
3 Diskussionen

Einladung: Informatik-Oberseminar Herr Aleksandar Mitrevski

von Maaßen, Leany

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 30.01.2023, 13:00-14:00 Uhr Ort: Informatikzentrum, Ahornstraße 55, Raum 5053.2 (B-IT-Hörsaal) Referent: Herr Aleksandar Mitrevski, M. Sc. LuFG Informatik 5 Thema: Skill Generalisation and Experience Acquisition for Predicting and Avoiding Execution Failures Abstract: For performing tasks in their target environments, autonomous robots usually execute and combine skills. Robot skills in general and learning-based skills in particular are usually designed so that flexible skill acquisition is possible, but without an explicit consideration of execution failures, the impact that failure analysis can have on the skill learning process, or the benefits of introspection for effective coexistence with humans. Particularly in human-centered environments, the ability to understand, explain, and appropriately react to failures can affect a robot's trustworthiness and, consequently, its overall acceptability. Thus, in this dissertation, we study the questions of how parameterised skills can be designed so that execution-level decisions are associated with semantic knowledge about the execution process, and how such knowledge can be utilised for avoiding and analysing execution failures. The first major segment of this work is dedicated to developing a representation for skill parameterisation whose objective is to improve the transparency of the skill parameterisation process and enable a semantic analysis of execution failures. We particularly develop a hybrid learning-based representation for parameterising skills, called an execution model, which combines qualitative success preconditions with a function that maps parameters to predicted execution success. The second major part of this work focuses on applications of the execution model representation to address different types of execution failures. We first present a diagnosis algorithm that, given parameters that have resulted in a failure, finds a failure hypothesis by searching for violations of the qualitative model, as well as an experience correction algorithm that uses the found hypothesis to identify parameters that are likely to correct the failure. Furthermore, we present an extension of execution models that allows multiple qualitative execution contexts to be considered so that context-specific execution failures can be avoided. Finally, to enable the avoidance of model generalisation failures, we propose an adaptive ontology-assisted strategy for execution model generalisation between object categories that aims to combine the benefits of model-based and data-driven methods; for this, information about category similarities as encoded in an ontology is integrated with outcomes of model generalisation attempts performed by a robot. The proposed methods are evaluated in multiple experiments performed with a Toyota Human Support Robot. The main contributions of this work include a formalisation of the skill parameterisation problem by considering execution failures as an integral part of the skill design and learning process, a demonstration of how a hybrid representation for parameterising skills can contribute towards improving the introspective properties of robot skills, as well as an extensive evaluation of the proposed methods in various experiments. We believe that this work constitutes a small first step towards more failure-aware robots that are suitable to be used in human-centered environments. Es laden ein: die Dozentinnen und Dozenten der Informatik _______________________________ Leany Maaßen RWTH Aachen University Lehrstuhl Informatik 5, LuFG Informatik 5 Prof. Dr. Stefan Decker, Prof. Dr. Matthias Jarke, Prof. Gerhard Lakemeyer Ph.D., JunProf. Dr. Sandra Geisler Ahornstrasse 55 D-52074 Aachen Tel: 0241-80-21509 Fax: 0241-80-22321 E-Mail: maassen(a)dbis.rwth-aachen.de

1 Jahr, 5 Monate

Einladung Informatik-Oberseminar Marcus Völker

von Völker, Marcus

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Montag, 06. März 2023, 14:00 Uhr Ort: Raum 2202 (HBau, 2. Stock), Ahornstr. 55 Hybrid über Zoom: https://rwth.zoom.us/j/99712637671?pwd=U0UzV1JiL1hwWnlzNW5pUC9hVDdOUT09 Referent: Marcus Völker, M.Sc. RWTH Informatik 11 Embedded Software Thema: Policy Iteration for Value Set Analysis of PLC Programs Abstract: Ensuring the correct behaviour of computing systems is an important task to prevent danger to the users of such systems. To do this, many analysis techniques have been developed that can find bugs in software, or prove that it complies to some specification of correct behaviour. Among these techniques, a fundamental analysis is value set analysis (VSA), which can determine an approximation of the program variables' values at each point of the program. This is very important information, as many faulty behaviours can be traced back to variables taking unexpected values, such as division by zero, access to uninitialised memory or outside a buffer, or unreachable code. While classically, value set analysis is performed with the algorithm of Kleene iteration, another approach called policy iteration has been developed in recent years that provides an alternative with the potential of finding similar or better results than Kleene iteration in less time. Policy iteration works by using a heuristic to simplify the program in a certain way, finding the value sets of that program, and then checking whether the result is applicable to the original program. If yes, the results are used, otherwise different simplifications have to be checked, until a usable result is found. As policy iteration is a heuristic algorithm, it makes certain assumptions about program behaviour in order to achieve good results. It turns out, however, that these assumptions are not guaranteed if the program contains errors which cause it to behave differently than expected. As we use program analysis to find errors such as this, assuming an error-free program is not necessarily a good assumption. In this thesis, we show several ways to improve the original heuristic by focusing on program loops. First, we present a way to use a pre-analysis to determine some aspect of the loops' behaviours, and use this information in order to build a heuristic that leads to more accurate solutions than the standard heuristic in many cases, at the cost of additional running time necessary to perform this pre-analysis. Then, we show a way to reinterpret branches as loops if they occur in cyclical code, which is typical for programs in reactive systems, such as the systems used for factory automation. This allows us to use our loop heuristic on a wider variety of programs, even though the cost becomes even greater, and it is useful only in specific cases. Afterwards, we show how to remove the pre-analysis to gain back the lost time, while still retaining similarly good results to the expensive version introduced before. This has the additional benefit of allowing usage of the algorithms on branches as in the second approach, without incurring any additional costs. We motivate that this is the version of policy iteration that should be used in general with an extensive evaluation of generated programs. Finally, we show a way to analyse polynomial inequalities with value set analysis by reinterpreting them as conjunctions of simpler inequalities. This not only allows us to improve value set analysis results on programs that feature such inequalities, but also makes these programs accessible to policy iteration. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 Jahr, 5 Monate

Einladung Informatik-Oberseminar Sascha Müller

von Noll, Thomas

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * +********************************************************************** Zeit: Donnerstag, 12. Januar 2023, 10:00 Uhr Ort: 9222 (Gebäude E3, Informatikzentrum) Referent: Sascha Müller M.Sc. DLR Braunschweig Thema: Synthesizing FDIR Recovery Strategies for Space Systems Abstract: This talk proposes an inherently non-deterministic model for Dynamic Fault Trees (DFTs) to analyze Fault Detection Isolation and Recovery concepts with a particular focus on the needs of space systems. Deterministic recovery strategies are synthesized by transforming these non-deterministic DFTs into Markov automata. From the corresponding scheduler, optimized to maximize a given RAMS metric, an optimal recovery strategy can then be derived and represented by a model we call recovery automaton. We discuss dedicated techniques for reducing the state space of this recovery automaton and investigate lifting the approach to a partially observable setting. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 Jahr, 5 Monate

2024

2023

2022

2021

2020

2019

2018

informatik-vortraege Januar 2023