Overview

Please find the schedule of GKmM Summer School 2010 below (as of 2010/08/12):

* TUD-VT UxV experiment is not an official part of the summer school.

Talks

Complex Event Detection and Notification in Mobile Sensor-Actor Networks (Alejandro Buchmann)

The behavior of sensor-actor networks is well described by Boyd’s OODA cycle. The Observe-Orient-Decide-Act cycle was developed by Boyd, a former fighter pilot, to describe the behavior of reactive systems. In the Observation phase primitive events are detected; in the Orientation phase they are combined, enriched with context information, and higher level events are derived; in the Decision phase the best course of action is determined based on the results of the Orientation phase; and in the Action phase decisions are played out yielding new events and situations.

The critical phase is the Orientation phase where complex events are derived and contextualized. This phase is particularly difficult in a distributed environment with mobility, potentially unstable communication and arbitrary delays. In this lecture we will cover the operators of event algebras, the semantics of different event consumption modes, mechanisms for combining infinite streams of events with finite relations, as well as, problems associated with synchronization, false positives and false negatives, event lifecycle, event dissemination and security.

Trajectory Optimization and Task Allocation for Multi-Vehicle Systems (Oskar von Stryk & Christian Reinl)

At its core, optimal task allocation and trajectory planning for cooperating vehicles is characterized by complex problems. A wide range of tasks, whose fulfillment significantly depends on physical motion dynamics, leads to a tight coupling of discrete structure and continuous dynamics in systems analysis as well as in optimal control.

In practical applications, either specific heuristics are used or the non-linear motion dynamics is considered under oversimplifying assumptions. Usually, existing approaches can therefore only limitedly be generalized or transferred to other vehicle classes and tasks.

In a continuously growing area of new applications for cooperative autonomous multi-vehicle systems, the development of broadly applicable methods is of particular significance. In modeling and in optimal planning, the corresponding concepts have to consider the system's basic characteristic to be applied not only in development of new control strategies but also in system design. A concept for modeling, approximation and optimization will be presented that is based on the theory of hybrid dynamical systems, on non-linear mixed-integer trajectory optimization and on model-predictive methods from control theory.

Using hierarchical hybrid automata allows for modeling the tight discrete-continuous coupling. By applying an appropriate transformation, the model is made accessible for mathematical optimization. In particular, linear approximations are used for reasons of numerical efficiency and global optimality of the resulting mixed-integer linear optimization problem.

Solving these discrete-continuous optimization problems allows to compute approximate solutions for various problems of cooperative behavior and can -- for special cases --  already be applied within decentralized real-time feedback control architectures in task allocation. For representative benchmark scenarios as well as for new problems -- like maintaining wireless communication among vehicles -- numerical results are presented, that are demonstrating the competitiveness of the proposed concepts and are sounding out their limits.

The presented methods allow for estimations in systems design and for reference solutions in development of heuristic controller mechanisms -- focusing on the core problem of cooperating vehicles, considering physical locomotion dynamics and the characterizing discrete-continuous coupling of states.