autotune: update README with fixed-wing rate tuning instructions

Author: mahima-yoga

autotune/README.md

@@ -29,20 +29,93 @@ python3 autotune.py
 
 ![image](https://github.com/user-attachments/assets/fcdf5c25-d92d-4487-9736-e77f6576d180)
 
+# Tuning Worflow
+
+## Fixed-Wing Rate Tuning 
+
+**Prerequisites**:
+
+- Requires valid airspeed measurement on vehicle
+- Correctly set FW_AIRSPD_TRIM
+
+### Flight Maneuvers
+
+Using default PX4 gains. Perform these maneuvers separately for each axis: roll, pitch, and yaw.
+
+1. Requires valid airspeed measurement, and a correctly set FW_AIRSPD_TRIM.
+2. Fly in stabilized mode and start from level flight.
+3. Apply a manual sine chirp input (sine wave with increasing frequency) on the selected axis. Avoid any other control inputs simultaneously.
+4. For pitch and yaw: The amplitude of the input should not be too high, but should be enough to overcome noise.
+5. Repeat the maneuver a few times if necessary to capture good data.
+
+### Using the Tool
+
+#### Load the flight log and select the relevant tuning loop (roll, pitch, or yaw):
+
 ![image](https://github.com/user-attachments/assets/f962a004-eae8-433f-9b7e-adc618dcbeae)
 
+#### Window selection:
+
+1. Select the portion of the log where the maneuver occurred.
+2. Ensure the window starts from level flight and covers the full maneuver.
+3. (Experimental) Use the coherence plot to check whether your input-output data have a strong enough relationship for frequency response identification. For fixed-wings:
+   i. We are interested in frequencies between 0.5-10Hz
+   ii. Values > 0.6 are sufficient
+   iii. If the coherence function is oscillating dramatically, the frequency response identification may not be reliable.
+4. Once you have identified the window, load the selection into the tool.
+
+<table>
+  <tr>
+    <td align="center"><b>Good window selection</b><br>
+      <img src="https://github.com/user-attachments/assets/25c30295-6825-4cb4-828e-47eaa17e1557" width="400"/>
+    </td>
+    <td align="center"><b>Bad window selection</b><br>
+      <img src="https://github.com/user-attachments/assets/1e05d411-37cf-42ab-8d6f-0cbbbb00921c" width="400"/>
+    </td>
+  </tr>
+</table>
+
+#### Model Verification
+
+1. Check that the estimated dynamic model correctly tracks the actual aircraft output (a).
+2. If not, adjust the system order (number of poles and zeros) and/or delay parameters (b).
+3. Verify that the system is minimum-phase (i.e., zeroes lie inside the unit circle) (c).
+
+TODO: update this image to include numbers and make the boxes the same colour
+![image](https://github.com/user-attachments/assets/1172f4a1-7d76-4ef5-8ad9-b0eb05b85a35)
+
+#### Tuning Gains
+
+1. Set all gains to zero
+2. Increase the P-gain until the step response shows oscillations. Then reduce the P-gain by approximately 50%.
+
+    <table>
+    <tr>
+        <td align="center"><b>Increase P-gain </b><br>
+        <img src="https://github.com/user-attachments/assets/db05a018-f9fb-4b3c-9f55-a4d7a577251f" width="400"/>
+        </td>
+        <td align="center"><b>Decrease by 50%</b><br>
+        <img src="https://github.com/user-attachments/assets/77dc38aa-f829-4878-a58e-32d775fa542c" width="400"/>
+        </td>
+    </tr>
+    </table>
+
+3. Adjust the I-gain to achieve a good step response with acceptable steady-state error.
+
+    ![image](https://github.com/user-attachments/assets/78f3bca5-2642-44e2-b3f2-7444fc4ffa40)
+
+4. Fine-tune both gains as needed. At the 1 second mark, a disturbance is injected into the system. The tuned response to this disturbance is visible in the step response plot.
 
-<img width="822" height="796" alt="Screenshot from 2025-08-20 15-21-50" src="https://github.com/user-attachments/assets/25c30295-6825-4cb4-828e-47eaa17e1557" />
+5. Use the Bode plot to confirm that the resonant frequency peak remains below 0dB.
 
-<img width="822" height="796" alt="Screenshot from 2025-08-20 15-22-15" src="https://github.com/user-attachments/assets/1e05d411-37cf-42ab-8d6f-0cbbbb00921c" />
+    ![image](https://github.com/user-attachments/assets/86d99295-43a8-47a9-b052-324ab5ac3082)
 
+### Applying the Gains
 
-<img width="1247" height="427" alt="Screenshot from 2025-08-20 15-36-18" src="https://github.com/user-attachments/assets/77dc38aa-f829-4878-a58e-32d775fa542c" />
-<img width="1247" height="427" alt="Screenshot from 2025-08-20 15-37-08" src="https://github.com/user-attachments/assets/78f3bca5-2642-44e2-b3f2-7444fc4ffa40" />
-<img width="1365" height="427" alt="Screenshot from 2025-08-20 15-39-30" src="https://github.com/user-attachments/assets/86d99295-43a8-47a9-b052-324ab5ac3082" />
-<img width="1720" height="538" alt="Screenshot from 2025-08-21 15-08-04" src="https://github.com/user-attachments/assets/2db77238-bd33-4454-af1f-a38901571e88" />
-<img width="877" height="756" alt="Screenshot from 2025-08-20 15-32-47" src="https://github.com/user-attachments/assets/1172f4a1-7d76-4ef5-8ad9-b0eb05b85a35" />
+1. The gains shown in the parallel form correspond to the PX4 rate controller parameters.
+   ![image](https://github.com/user-attachments/assets/2db77238-bd33-4454-af1f-a38901571e88)
 
-<img width="1247" height="427" alt="Screenshot from 2025-08-20 15-35-47" src="https://github.com/user-attachments/assets/db05a018-f9fb-4b3c-9f55-a4d7a577251f" />
+2. If the identified parameters are strongly different to the currently set ones (more than 30% for multiple axes), an incremental application while flying is recommended (not prior to takeoff, during takeoff, or during landing).
 
+3. For all loops tuned with the strategy outlined in this guide: increase the respective integrator limit parameters to 1 (FW_RR_IMAX, FW_PR_IMAX, FW_YR_IMAX)