*7.2. Test 2: Sine Inputs*

The next test feeds a more complex signal composed of a sine wave with increasing frequency instead of a step signal, as shown in Figure 14. The test is conducted in order to validate the models for simple dynamic movements. As mentioned before, these cases are not explicitly captured by the FSS or training data. Figures 15 and 16 show the outputs.

The Set Membership followed the output of the system closer than in the previous test. It scored 93.0012% for pitch and 93.8389% for roll, much closer to the validation results. It is important to mention that the SM follows the dynamics and does not have the bias error observed in Test 1. Therefore, the FSS properly captures the continuous dynamic behavior, but it requires additional information to capture static behaviors, as required in Test 1. This is true also for the NN model, which also improved. It scored 98.1058% in pitch and 98.3087% in roll. The RLS roll output shows the disadvantages of this model. Due to the decoupling of the signal, while the initial movement causes little movement on the roll axis, RLS cannot process them and scores a worse fit than in the rest of the test. The fit scores are 78.4813% in pitch and 48.5090% in roll.

**Figure 14.** Test 2 input signal.

**Figure 15.** Test 2 results for pitch.

**Figure 16.** Test 2 results for roll.
