The process of automation has affected the workforce since the advent of the steam engine.
I agree with you entire post, but i like to point something out that seems from my current viewpoint not entirely perfect. And just for the sake of my autism i will try to point this out, despite my perception from your post that you understand quite well this topic.
For the automation process you could go back to the water and wind mills, but if you analyze what is the core point of this inventions, then you will discover that it is about the use of energy that is not part of the human body and then you understand this as an energy investment by reducing the necessary human work to achive a certain goal or execute a certain task. And then you can go as far back as domestication of animals and the use of their labor in agrecultural and building sector to conserve 'human' energy or work time and access the stored sunlight energy in for humans undigestable plant form.
And now this gets quite clear where and when automatisation processes can take place.
Achieving fully automated planning and control of an industrial process is a difficult problem.
Yes, but if you take my previous post into consideration, then you will recognise that the first work will be automatised that requires a lot of human capital or that is quite repetitive (here are two dimenstions: intesity and variability). The variability and the control over the task is what is currently hindering the total robotisation of the society.
Take as an example the robotisation in nursing of the old people: Because of the lack of control over the process the inventor of the human lifting apparatus has strangled himself in his own invention.
But due to a rising of computing power and minaturisation of sensors with sensorfusion we will achive the highly desired goal of lifting and turning the old people in their beds by an apparatus in near future. Without killing or breaking them.
So the only point remains is the task flexible 'roboter'. There are several approaches towards this problem and not all give all the desired solutions, but i suggest that machine learning and world simulations can solve this task in our lifetime. (About this i could write entire essays, so i will elaborate only if desired.)
Blue color worker will stay as long as we cannot adapt the environment towards the roboter, if we cannot adapt the roboter towards the environment.
White color are also in danger of being rationalised away at least to some degree (see the first part). Even the scientists are in danger and this began as the physicists began to use neural nets to analyze particle collisions pictures (because of the very high amount of taken pictures).
The last part i have cut short, because i have to do something and i cannot concentrate.