Much of the meat we enjoy to eat, undergoes a curing process for preservation, tenderizing, flavoring, and presentation purposes. Curing on an industrial scale is commonly done by injecting brine, a solution of salt and other ingredients, into the meat. Current methods rely on manual control of process settings. Natural variations, between and within meat pieces, are ignored resulting in sub-optimal dosing and distribution of brine. Were the process parameters instead to be adapted to the individual meat piece and the specific area, yield and quality could be improved. This paper reports research to investigate the fundamental aspects of the injection process, constructs a process model for existing machines and proposes a self-calibrating controller based on Reinforcement Learning. A vision based robotic injection system is presented for experimentation with the injection process, with the specific purpose of determining the potential for adapting process parameters to the natural variation in the meat pieces.
