Kendali Navigasi Lateral Berbasis Model Predictive Control pada Pesawat Tanpa Awak Sayap Tetap untuk Misi Loitering

Zaky Nur Ihsan, Prof. Dr. Ir. Jazi Eko Istiyanto, M.Sc; Dr. Andi Dharmawan, S.Si., M.Cs.

2026 | Tesis | MAGISTER ELEKTRONIKA DAN INSTRUMENTASI

Misi loitering pada fixed-wing Unmanned Aerial Vehicle (UAV) membutuhkan kemampuan pelacakan lintasan yang stabil dan presisi, terutama ketika wahana mengalami gangguan lingkungan seperti angin dan turbulensi. Metode navigasi berbasis L1 guidance law yang umum digunakan bersifat reaktif karena koreksi lintasan dilakukan setelah cross-track error muncul, sehingga performanya menurun pada kondisi gangguan dinamis. Penelitian ini mengusulkan kendali navigasi lateral berbasis Model Predictive Control (MPC) untuk meningkatkan akurasi dan stabilitas lintasan loitering UAV sayap tetap. Kendali dirancang menggunakan model lateral UAV yang mencakup posisi horizontal, heading, dan sudut roll. MPC digunakan sebagai outer-loop controller untuk menghasilkan referensi sudut roll optimal berdasarkan prediksi perilaku sistem pada prediction horizon, sedangkan Linear Quadratic Regulator (LQR) digunakan sebagai inner-loop controller untuk menjaga kestabilan sikap roll. Fungsi biaya MPC mempertimbangkan cross-track error, heading error, perubahan input kendali, serta batasan fisik UAV berupa sudut bank maksimum dan laju perubahan aktuator. Implementasi dan pengujian dilakukan pada lingkungan Software-In-The-Loop (SITL) yang terintegrasi dengan Gazebo dan komunikasi MAVLink. Hasil pengujian SITL menunjukkan bahwa MPC mampu menghasilkan lintasan loitering yang lebih akurat dan tahan terhadap gangguan angin dibandingkan L1 dan PID. Pada kondisi angin steady 6 m/s, kendali MPC menghasilkan IAE cross-track error sebesar 58,820 (m·s) dan RMS sebesar 0,875 (m), dibandingkan L1 sebesar 344,483 (m·s) dan 4,630 (m), serta PID sebesar 283,624 (m·s) dan 4,007 (m). Selain meningkatkan akurasi lintasan, MPC menghasilkan respons roll yang lebih halus dengan bank-rate RMS sebesar 12,133 derajat per detik  lebih rendah dari L1 (24,974 derajat per detik) maupun PID (38,445 derajat per detik). Evaluasi kinerja komputasi pada Jetson Nano menunjukkan rata-rata waktu eksekusi sebesar 39,49 (ms) dengan nilai maksimum 87,01 (ms), masih di bawah budget siklus kendali 100 (ms), sehingga kendali ini layak untuk implementasi real-time pada UAV sayap tetap.

Loitering missions in fixed-wing Unmanned Aerial Vehicles (UAVs) require stable and accurate path-following performance, particularly under environmental disturbances such as wind and turbulence. Conventional navigation methods based on the L1 guidance law are inherently reactive because trajectory corrections are applied only after cross-track error occurs, resulting in degraded performance under dynamic disturbances. This research proposes a lateral navigation controller based on Model Predictive Control (MPC) to improve the accuracy and stability of fixed-wing UAV loitering trajectories. The proposed controller is developed using a lateral UAV model consisting of horizontal position, heading, and roll dynamics. MPC is implemented as an outer-loop controller to generate optimal roll-angle references by predicting future system behavior over a prediction horizon, while a Linear Quadratic Regulator (LQR) is employed as an inner-loop controller to maintain roll attitude stability. The MPC cost function incorporates cross-track error, heading error, control effort variation, and physical UAV constraints, including maximum bank angle and actuator rate limits. The controller is implemented and evaluated in a Software-In-The-Loop (SITL) environment integrated with Gazebo and MAVLink communication. The SITL evaluation results demonstrate that the proposed MPC produces more accurate, disturbance-resilient, and smoother loitering trajectories than both the conventional L1 guidance law and a PID controller.. Under steady 6 m/s crosswind, the MPC controller produces a IAE cross-track error of 58,820 (m·s) and an RMS of 0,875  (m), compared to 344,483 (m·s) and 4,630 (m) for L1, and 283,624 (m·s) and 4,007 (m) for PID. In addition to improved tracking accuracy, MPC yields a smoother roll response with a bank-rate RMS of 12,133 degrees per second lower than both L1 (24,974 degrees per second) and PID (38,445 degrees per second). Computational performance evaluation on a Jetson Nano shows a mean execution time of 39,49 (ms) with a maximum of 87,01 (ms), remaining below the 100 (ms) control-cycle budget, confirming the controller's suitability for real-time implementation on fixed-wing UAVs.

Kata Kunci : UAV, Linear Quadratic Regulator, MPC , loiter, Full-state-feedback