As a controls engineer, building automated manufacturing equipment, the general principle that we've found to work best is to let the humans do human things and

As a controls engineer, building automated manufacturing equipment, the general principle that we've found to work best is to let the humans do human things and the machines do machine things. A person with a spoon, tediously and laboriously stirring ingredients in a pot, is a poor way to make use of that person's intelligence, creativity, and flexibility. An electric motor just does the job better. On the other hand, by the Anna Karenina principle, cheese inspection is one of those tasks where there are a thousand unique and unexpected ways for a cheese to be wrong but only one way for it to be right. It's very hard to design an inspection that would catch everything and miss nothing that a human would trivially see, smell, or feel, while also minimizing false positives. The robotic wheel flipper is somewhere in the middle: humans are great at navigating complex environments, and while you can design a uniform, controlled environment that a complex AMR can navigate, and space the wheels out regularly, it seems like half the task (rotating the cheese) is something ideally suited for a robot arm and half the task (getting to the cheese) is something better suited for a human. Humans can maintain the environment and debug the process, the robots can flip the cheese. With respect to the cheese testing, I think a good middle ground is tool-assisted human inspection. Instead of/in addition to a hammer, give them an ultrasonic transducer and audio analysis toolset. Let them manipulate the cheese, but also give them objective numerical data on the frequency response and calculated porosity. It's easy for a person to recognize when a cheese is more hollow-sounding than the previous, but the first cheese of the day might be hard to for a human to recognize, and better tools than a primitive hammer can help with that.