Figure 7 | Scientific Reports

Figure 7

From: Reinforcement learning optimisation for graded metamaterial design using a physical-based constraint on the state representation and action space

Figure 7

(a) Schematic representation of the decisions of the RL agent according to the previously proposed interpolation rule22 and (b) the current approach. The previous proposal made use of B-spline interpolation. In the current proposal, the interpolation points, depicted as stars, are placed at the start, at the end, at 1/2 of the resonator array. The first agent action places just the first resonator. The second action places all the resonators by setting the length of the last resonator and enforcing a linear grading rule. The third action sets the position of the middle interpolation point. Finally, the resonators’ lengths are set by the parabola passing through the three interpolation points.

Back to article page