I'm afraid your approach to this topic is wrong. You are interested in "a more scientific approach", but a dog is not a machine.
In science, chemical A + chemical B equals chemical C.
In dog training, dog A + training method B must not always equal behavior C. Every dog is an individual with an own personality.
As far as I know, people who train dogs to herd have years of experience and have to adapt the training to the individual personality of each dog. That's why you find more general information and anecdotal stories than scientific training manuals online.
As to the precise questions you asked:
- Herding training usually starts when the puppies are a few weeks old. They don't have to do any work at that age, but they have to be socialized with the animals they're supposed to herd. If you want your dog to herd ducks, it has to be familiar with ducks as "friends" as opposed to "prey".
- The puppies are taught basic commands like "sit" and "stay" with positive reinforcement (reward wanted behavior, ignore unwanted behavior).
- Once they know the basic commands, they're taught more complex commands for herding like "go left / right", "go further away" and "come closer", again with positive reinforcement. Most people use whistles instead of spoken commands and use different commands for each dog.
- The young dogs are often taken to the herds to watch adult dogs herding. They are given very short and simple tasks at first. Herding is a very strenuous activity, both physically and mentally, and the young dogs have to adapt to more complex tasks slowly.
The dogs must be absolutely obedient before each further step in the training. They become the remote control tool of the shepherd, but the shepherd cannot control the herd if they cannot control their dog/s.
The training must be done very regularily (at least twice a week) for a long time (I think at least half a year, probably longer). That's one of the reasons why many shepherds buy trained herding dogs instead of training them themselves and why those trained dogs are extremely expensive.
Positive reinforcement (only rewarding good behavior and ignoring wrong behavior) generally works better than negative reinforcement (punishing wrong behavior) because the dogs need to have fun while doing their job. If a dog is afraid of punishment, it'll rather do nothing than risk doing the wrong thing.