Communications in Embedded Control Systems - Project Summary This project investigates how different embedded software and hardware components should communicate with each other in order for the overall, global system to behave in a satisfactory way. The main area of investigation is the information theoretic content of the transmitted signals, i.e. a study of how many bits of information need to be transmitted between the different components in order to make the physical system meet its specifications. For instance, autonomous robotic systems are normally relaying on a variety of heterogeneous sensors, and the question then becomes, which of the sensors are needed? And furthermore, is it possible to compress the data in a systematic way, e.g. using virtual sensors, so that information that is not essential to the current task, or mode of operation, can be discarded?
Since the sensory data needed for accomplishing a certain task depends both on the control law, the dynamics of the system, and the complexity of the task, information theory provides a tool for treating questions concerning sensor and actuator selection and control mode design in a unified way. This project is focused on the following two enabling areas of research: 1. Modeling of the information theoretic content of the symbolic, computer generated inputs used for controlling continuous, mechanical devices. 2. Selection of embedded sensors and actuators for a multi-agent, robotic system. This selection is done so that the control system meets its specifications in the presence of bandwidth constraints, i.e. so the robots maintain formation while avoiding obstacles in the environment.
In this project computer generated, or linguistic control commands are modeled as strings of mode descriptions that can be read by trigger based hybrid systems. In other words, models for software based control of physical, continuous devices are developed in such a way that the actual control signals have a clear information theoretic content in terms of code lengths, thus connecting the field of multi-modal control with that of information theory. This model enables the investigation of component integration issues. Since each sensor or actuator contributes to the total number of bits needed for coding the control signals, by minimizing the code lengths subject to the constraint that the system should behave in a satisfactory way, a natural measure of what components are needed in the current mode of operation is derived. Furthermore, the explicit focus on code lengths has immediate implication on coding theory, i.e. a theory for how the information shared between different embedded components should be represented is a direct consequence of this research.
The information theoretic approach to control theory under investigation in this project has a number of potential applications outside the field of embedded control. For instance, in teleoperated robotics the control signals are transmitted over communication channels in which the presence of channel noise makes it preferable to transmit instructions that are as short as possible. A related problem arises in the area of minimum attention control, where an attention functional is defined as a measure of the control variability.
