Questions about digital design using MCU/FPGA vs reinforcement learning

[bhl777]

Hi All,
I was reading some of the basic introduction to the reinforcement learning algorithm, and it seems to me that if the algorithm applies to the hardware, it is similar to the control loop tuning in digital design.
Would anyone advise
(1) if they are the same thing in principle if we talk about the hardware application
(2) what are the key differences between digital control design and reinforcement learning?

Thank you!

 

[BradtheRad]

For general purpose they're similar in principle.

A computer program might decide when to fire up the boiler that heats a building's water. The goal is to maintain a certain temperature, and the heat must be on a certain amount of time to make the temperature rise. Such a system can be programmed to learn (through a number of cycles) how to time On-Off periods so that the boiler doesn't get overly hot or cool.

Is that the sort of thing your question is about? The digital system has a control loop, and learns. I'm not certain about it being a reinforcement learning algorithm. There's Hexapawn Educable Robot (invented by Martin Gardner), where you reinforce it to play an unbeatable game, by removing the token representing its previous losing move, and when it wins you add a token representing its previous move. In other words, rewarding and punishing to instill reinforcement learning.

 

[bhl777]

Thank you BradtheRad! Would you also provide some suggestions to quickly learn digital design and embedded programming?

My background is more related to analog electronics. Sometimes I found it difficult to understand some of the C code because they are related to the CPU operation (like interruption). Would you give me some suggestions to quickly ramp up the study, like the online course/book or anything else?

Thank you!

 

[BradtheRad]

Microcontrollers (or Arduino, Raspberry, and the like) have become popular and inexpensive devices. Quickest way to learn is to start with examples provided in the package, then to adapt these by writing your own code. Get familiar with the entire set of command words. The examples show how to handle interrupts (usually button presses, keypresses, mouse clicks, etc.) Online groups are a resource as well.

A big hurdle is to make your device interact with the outside world. Many projects take input via the ADC terminal (Analog-to-Digital Converter). Or communicate with other devices by serial protocol. It's essential to use compatible volt levels. And to use correct syntax. Often trial-and-error is the means to success.