Modeling and simulation plays a central role in almost all disciplines of science and engineering. In computer science, modeling and simulation is central for the development of autonomous, concurrent, self-organizing systems. Methods and tools have to be developed to account for the requirements of these diverse applications. The lecture provides an overview of fundamental methods and techniques used for modeling, efficient execution of simulation models, and the design of experiments.

Extract from the content:

• Dynamic systems: discrete-stepwise, discrete-event-based, and continuous
• Simulation models formally: syntax and semantics
• Modeling approaches: cellular automata, discrete-event-oriented system specification (DEVS), and hybrid automata
• Dealing with uncertainty: stochastic Petri nets and process algebras
• Efficient execution: data structures and parallel, distributed simulation algorithms
• Experiment design: input and output analysis

Further information: 3V + 1Ü, 6LP, SS

Parallel and distributed simulation methods are of particular interest when simulating complex and large systems, e.g., to evaluate routing protocols in networks with millions of nodes, to monitor air traffic online, or to predict the spread of epidemics. This module provides knowledge of parallel, distributed algorithms for executing discrete event models efficiently.

Extract from the content:

• Null message algorithm
• Deadlock detection and recovery
• Time warp algorithm
• Global virtual time and transient messages
• Incremental state saving
• Reverse computation
• Exploiting GPUs

Further information: 3L + 1E, 6 LP, SS

The lecture is centered around the question of how software agents can make decisions about goal-oriented actions in dynamic environments. The lecture presents methods from artificial intelligence such as knowledge representation (e.g., modal logics) and planning (e.g., distributed plans), and methods from machine learning, here in particular reinforcement learning, as well as concepts from related areas, for example linguistics (speech act) and game theory (e.g., Nash equilibrium, and Pareto optimum), and discusses their significance for the design of autonomous, intelligent software agents. A project accompanies the lecture to deepen the understanding of theoretical concepts by a practical application. 

Extract from the content:

• Beliefs Desires Intentions: architecture of deliberative agents
• He knows that he knows not: the role of modal logic
• Markov decision processes: reinforcement learning for multi-agent systems
• The prisoner's dilemma and equilibria: game theory
• Communication: from speech acts to ACL
• Negotiation protocols: between consensus and cheating
• Reputation: own and second-hand experiences 
• Planning: Distributed plan generation or execution

Futher information: 3V + 1Ü, 6 LP, WS

Data play a central role in modeling and simulation.
To calibrate and validate a simulation model, a multitude of different simulation experiments can be executed which rely on diverse data.
At the same time, these simulation experiments may reveal important information about the data.
The lecture gives an overview about experiment design methods, data analysis methods, and about different types of simulation experiments, including sensitivity analysis, statistical model checking, optimization, parameter estimation and uncertainty quantification.
The lecture includes a student project in which a simulation model shall be re-developed from literature.
The developed simulation model as well as the data shall be extensively probed by analysis and extensive simulation experiments.

Extract from the content:

• Work smarter not harder: experiment design methods
• SESSL: a domain-specific language for specifying and executing simulation experiments in Scala
• Optimization: more than hill climbing
• Making hypotheses explicit: the virtue of statistical model checking
• Bayes: statistical parameter estimation and uncertainty quantification

Further information: 2L + 2E, 6 LP, WS