Solution 1 was a classic PID. The pendulum swung, paused, then crashed. Solution 2 added feed-forward. It worked in simulation, but the real hardware hummed with a chaotic tremor. Solution 3 used a lead-lag compensator. Better, but the wind knocked it over every time. Solution 4 was state feedback. Elegant, but her gains were too aggressive. The motor screamed. Solution 5—LQR. Perfect on paper. In the lab, the cart twitched like a dying insect. Solution 6 was adaptive. The code was beautiful. The hardware caught fire.
“No one uses that,” she whispered. “Too much math.” control systems engineering 8th solution
She had tried seven solutions.
Now, at 2:00 AM, the lab smelled of burnt resistors and desperation. Her professor, Dr. Hsu, had a rule: “You get eight attempts. After that, the pendulum wins.” Solution 1 was a classic PID
Dr. Elara Vance stared at the blinking cursor on her terminal. Above it, the assignment title glowed like a dare: “Design a stabilizing controller for the inverted pendulum on a cart. Non-linear friction present. Wind disturbance modeled as Appendix F.” It worked in simulation, but the real hardware
She flipped to Chapter 13. Midway through a derivation, she noticed a handwritten note in the margin—left by a previous owner. It read: “Forget continuous time. Sample at 0.05s, then solve for z-domain pole placement with a Smith predictor. The 8th solution is the one you write yourself.”