Published February 13, 2023 | Version 1.0
Thesis Open

Optimal sensor placement for active flow control with deep reinforcement learning

Creators

Contributors

Supervisor:

Description

Report on investigation of 3 optimal sensor placement methods for active flow control of the fluidic pinball with deep reinforcement learning. Results have shown that the attention mechanism effectively reduces the number of sensors from 476 to 7 without losing performance, thereby identifying optimal sensors directly in the DRL optimization loop. 

Files

Optimal_sensor_placement_for_active_flow_control_with_deep_reinforcement_learning.pdf