Hybrid control of autonomous system

Researchers: Anders Lindquist, Xiaoming Hu, Magnus Egerstedt, Henrik Rehbinder, Claudio Altafini, Torvald Ersson and Petter Ögren.

Sponsor: The Swedish Foundation for Strategic Researches (SSF).

This project is part of the ongoing research effort of the Center for Autonomous Systems. In this project we use a hybrid system approach to study some issues concerning the integration of control and sensor systems for a mobile manipulator. In particular, the following issues are studied.

1.
Modeling of basic behaviors as dynamic systems. In order to take advantage of HDS methods, one should model individual behaviors of a robot (action responses to sensory inputs) as dynamical systems, or as a set of differential or/and difference equations. Then one can study problems such as path planning and obstacle avoidance for systems with possibly nonholonomic constraints.
2.
Modeling of tasks as hybrid dynamic systems. In a Behavior Based robot architecture, many behaviors are affecting the system simultaneously. Therefore questions concerning safety, task-achievement and liveliness (non-blocking behaviors) are hard to answer due to the complexity of the system. However, if these behaviors could be modeled as nodes in a hybrid automata, features about the system could be proved.
3.
Sensor fusion. The multiple sensor problem, known as sensor fusion, has been studied extensively, but a fully satisfying solution has not yet been provided and many research issues are still open. We study the basic problems of designing and tailor-making filter algorithms for our applications as well as more general filtering problems for sensors in distributed control systems.
4.
Feedback Control of Hybrid Systems. A few different problems on the control of hybrid system will be studied. By the use of abstraction, it is possible to simplify complex decision-making problems into hierarchical control systems. It is then necessary to have consistent methods for aggregating system properties from one level to another. This crucial problem in the design of complex control systems have only recently been addressed. We will study this problem together with the related problem of composition of hybrid systems.

Research 1999/2000
Geometric theory of linear stochastic systems
Integral quadratic constraints
Last update: February 21, 2001 by Anders Forsgren, anders.forsgren@math.kth.se.