Model-based Synthesis of Adaptive and Trustworthy

(ESR10) Model-based Synthesis of Adaptive and Trustworthy Control Software
Application area
Robotics
Host
Fortiss GmbH, Germany
Supervisor
Dr. H. Ruess
www.fortiss.org
Academic host
Technische Universität München, Germany
www.tu-muenchen.de
1st secondment
Eindhoven University of Technology, Netherlands
2nd secondment
Technolution, Netherlands
Project description
This ESR will consider a model-based approach for the correct-by-construction synthesis of control
software. The idea will be demonstrated considering model-based control of a robotic system.
State-of-the-art
In practice, robots have only limited resources (e.g. energy) available and control problems have
associated performance (e.g. guaranteed response time, possibly probabilistic) and safety constraints
(e.g. collision avoidance). The lack of verifiability and constraints on control performance restrict the
use of robotic solutions in many safety-critical domains. Moreover, in this context, the prevailing
sequential programming paradigm is not only resource-intensive but also results in rather inflexible,
costly, and incorrect solutions.
Novel contributions
The ESR will address robot control as a two-player game between the robot and its physical
environment. The robot wins the game if there is a control strategy for accomplishing all given goals
in an environment-tolerant way. Winning strategies will be synthesised by developing new
algorithms to solve Exists-Forall Satisfiability Modulo Theories (EFSMT) problems with a tradeoff
between precision and the resources used to solve these problems. Since environment models and
robot capabilities may change over time, the ESR will look at how control strategies are adapted
correspondingly.