Distributed control of multiscale microscopic chemical sensor networks

Micromachines and nanoscale devices can act together for high-resolution monitoring and action on multiple size scales within physical materials and biological organisms. Such multiscale systems require robust distributed controls responsive to heterogeneous and poorly characterized microenvironments. To develop such controls, this paper extends the partially observable Markov decision process formalism to machines operating asynchronously at multiple scales and with delays. We show how an approximation to this formalism readily provides aggregate performance measures needed for designing distributed controls. This approach identifies the aggregate behavior as arising from differential-delay equations approximating the system's dynamics. We illustrate this approach to distributed controls for chemical sensor networks consisting of micro-and nanoscale devices in the context of high-resolution biomedical diagnostics and treatment through localized drug delivery.