I have a system with 3 pumps connected to a water pressure sensor. The other side of the sensor connects to 7 water pipes with random and varying usage. The pumps have adjustable opening degrees from 0 to 100. The water pressure sensor output ranges from 0 to 1. Each of the 7 pipes has a random usage ranging from 0 to 300. I want to design an observation space and Gym environment for reinforcement learning to control the pumps so that the pressure stabilizes around 0.5. How should I design the observation space? Should the observation include: Current pressure (0-1) Usage of each of the 7 pipes (0-300 each) And what about the action space? I plan to control the opening degree of each pump. Could you suggest a proper way to design the Gym environment for this problem, including the observation space, action space, and reward function ideas? Continue reading...
