Distributed computing systems involving multiple sensors and actuators are everywhere. They are deployed  in manufacturing systems, large scientific apparatus, surveillance systems, sensor networks, systems for managing large-scale networks and data centers.

Today, each such environment develops its own control system. This system collects data from sensors, performs real-time analysis of collected data, archives data for off-line historical analysis, as well as controls and reconfigures the operations of the system at various time-scales. These control systems are often built for specific domains and deployments; consequently, they are inflexible and unable to evolve. Most changes in the underlying system deployments result in significant engineering effort to update the control system. Further, there is little reuse of these control systems across domains.

In this exploratory project, we plan to ask the following fundamental questions:

• Is it possible to design an “operating system” for a general class of multi-sensor-actuator systems?

• What set of abstractions should such an operating system support?

• How can the design of this operating system be made both platform- and domain-independent?

• How difficult would it be to create custom deployments of the control system for specific platform and domain?

• If we succeed in answering these questions, we can dramatically simplify, and thereby disrupt, the field of designing control systems for multi-sensor-actuator environments.

We plan to drive our explorations by gaining first-hand experience in designing control systems for two large-scale multi-sensor-actuator systems:
(1) The International Thermo-nuclear Experimental Reactor (ITER) [in collaboration with the Indian Plasma Research (IPR) Institute, Ahmadabad];
 (2) The Giant Meter-wave Radio Telescope (GMRT) [in collaboration with the National Center for Radio Astrophysics (NCRA), TIFR, Pune].