Teaching

If you are interested in a topic, or want to propse a topic in the area of Runtime Monitoring, Cyber-Physical Systems, Software Engineering or Learning Analytics, please contact michael[dot]vierhauser[at]uibk[dot]ac[dot]at

__________________________________________________

Extending PDDL with Quality Preference/Attributes

PDDL, the Planning Domain Definition Language, is a computer-readable language used in the field of artificial intelligence and automated robotic planning. It is specifically designed for defining the details of planning problems and domains in the context of automated planning and scheduling systems. PDDL provides a formal and expressive way to describe the various components of a planning problem, such as the initial state, the goal state, and the actions that can be taken by an agent or a robot to transition from one state to another. It allows the specification of the problem's characteristics, constraints, and the relationships between different elements.

However, particularly for successful human-robot collaboration, it is vital that humans can both express their requirements and they can comprehend the decisions that robots take. Requirements in this context are often related to quality objectives, such as performance, energy consumption, availability, or security. These objectives then have to be considered during automated planning, so that an optimal mission for the given concerns can be found. To help humans express their preferences for automated planning, the need for interactive approaches has emerged, guiding users through the planning process and providing timely feedback on different choices they make. 

The goal of this Master Thesis is to first get an overview of the existing capabilities of PDDL and other planning languages and in a second step extend PDDL so that it can incorporate the definition of Quality Attributes and respective user preferences.

__________________________________________________

Dynamic/Iterative Planning for Robotic Systems

Mission planning is a fundamental problem in mobile robotics. Domain-independent planners and the PDDL language, provide a standard and flexible framework for solving such planning problems. However, with the increasing size of the planning problem (e.g., a large number of nodes and edges part of a map), creating such a plan can be both resource and time-consuming. Waypoints and paths a robot can travel, for example in a warehouse, can be represented as a graph with constituent nodes (representing locations) and edges (representing the path).  Instead of fully creating a plan upfront, subsets of plans can be generated iteratively, and the robot can re-plan its path as it progresses in its mission.

The goal of this Master Thesis is to first get an overview of existing dynamic robot planning approaches,  and how the concept of cliques in graphs can be leveraged to divide the planning problem into smaller sub-problems. Based on this, a planning approach should be developed to create first inter-clique plans for course navigation, and subsequently intra-clique plans for detailed navigation.

__________________________________________________

Integrating Humans in Runtime Monitoring and Analysis of Cyber-Physical Systems

Cyber-Physical Systems have become increasingly pervasive in our everyday lives, with examples ranging from self-driving cars and autonomous vehicles to small unmanned aerial systems (or drones), used for delivery, surveillance, and rescue missions, to Cyber-Physical Production Systems with robots on a shop floor.
In this context,  Runtime Monitoring has been introduced as a means to collect and analyze sensor data and check the adherence of a Cyber-Physical System to both functional and non-functional requirements to, for example, ensure safe operation.

One aspect, however, that is often overlooked or neglected is the active and passive involvement of humans, e.g., controlling or interacting with certain parts of a CPS and their involvement in the monitoring process. Particularly with the new paradigm of  Human-Machine-Teaming, where humans are more closely interacting, and actively collaborating with CPS, collecting and analyzing information about a human actor engaged with a system could provide additional benefits, to both the system and humans alike.

The goal of this work is to first gain an overview of existing Monitoring Solutions for CPS, and subsequently develop concepts and ideas on how humans can be monitored, what data can be collected, and how data can be processed while taking into account privacy concerns.

Status: AVAILABLE

__________________________________________________

A Unified Language and Model for Defining and Managing Learning Competences 

Competency-based education has gained widespread popularity in recent years and has become increasingly important, particularly in higher education. Competencies are skills and abilities students are expected to master after they have successfully completed a course.
However, up until now, there is little to no tool support for defining and managing competencies, linking them to the respective course material and example questions.
The goal of this work is to first gain an overview of existing Competency-based education tools and methods and subsequently develop a model and language for describing and managing competencies and associated resources.

Status: AVAILABLE