Cognitive safety control for human-machine interaction

In order to introduce cognition, i.e. perception and autonomously reconfigured control, into manufacturing systems for increasing the flexibility and adaptivity this paper proposes an architecture of a cognitive safety controller (CSC). The CSC consists of a strategic controller (SC) and an operational controller (OC): the SC iteratively computes (sub-)optimal control strategies taking dynamic models, safety specifications, and learned knowledge into account. The operational controller (OC) provides certified safety by safe realization of the strategy computed by the SC. The paper illustrates the application of the CSC to a cognitive manufacturing system (CMS) in which a robot and a human operator share a workspace. Important components of this application are a procedure to compute (sub)optimal strategies for varying control specifications online and a scheme for implementing the strategy on a programmable logic controller (PLC).