NMPC-based Unified Posture Manipulation and Thrust Vectoring for Agile and Fault-Tolerant Flight of a Morphing Aerial Robot

TL;DR

The work tackles the challenge of enabling agile, fault-tolerant flight for a morphing aerial robot by developing a unified nonlinear model predictive control (NMPC) framework that simultaneously leverages posture manipulation and thrust vectoring. A reduced-order prediction model is embedded within a high-fidelity Simscape Multibody simulation to enable online trajectory optimization under actuator faults and aggressive maneuvers, without explicit fault diagnosis. The key contributions include (i) a dual NMPC configuration for fault-tolerant recovery and for agile trajectory tracking, (ii) a validated Simscape-based digital twin with ground-contact modeling and aerodynamic effects, and (iii) demonstration of robust fault recovery and high-speed turning capabilities, including turns up to $120^ op{}$ at speeds above $10 ext{ m/s}$. This framework advances resilient autonomous flight for multimodal morphing robots, with potential applications in planetary exploration and search-and-rescue in cluttered or GPS-denied environments. The results indicate that exploiting both thrust vectoring and joint articulation within a single optimization framework can maintain stability and performance under severe faults, reducing the need for fault detection and controller switching in real time.

Abstract

This thesis presents a unified control framework for agile and fault-tolerant flight of the Multi-Modal Mobility Morphobot (M4) in aerial mode. The M4 robot is capable of transitioning between ground and aerial locomotion. The articulated legs enable more dynamic maneuvers than a standard quadrotor platform. A nonlinear model predictive control (NMPC) approach is developed to simultaneously plan posture manipulation and thrust vectoring actions, allowing the robot to execute sharp turns and dynamic flight trajectories. The framework integrates an agile and fault-tolerant control logic that enables precise tracking under aggressive maneuvers while compensating for actuator failures, ensuring continued operation without significant performance degradation. Simulation results validate the effectiveness of the proposed method, demonstrating accurate trajectory tracking and robust recovery from faults, contributing to resilient autonomous flight in complex environments.

Figures (20)

• Figure 1.1: Various examples of aerial robotic systems.nan2022nonlineargabriele2023wang2023hao2022faultyao2024ke2023uniformshen2021mazare2024robustnguyen2019oconnell2024learningyu2024faultahmadi2023
• Figure 1.2: The M4 robot in rover, segway, and aerial configurations filip.
• Figure 3.3: Block diagram of the integrated control and simulation environment, highlighting the feedback loop between the NMPC optimizer and the Simscape multibody plant.
• Figure 3.4: Detailed schematic of leg articulation subsystem: sagittal and frontal joints per limb, STL parts assembly, and placement of thrust and external force blocks. Contact forces are applied between adjacent wheels to enforce physical realism and avoid limb interference.
• Figure 3.5: Implementation of aerodynamic drag and thrust-induced moment simulation within the Simulink environment. Drag torques are applied proportionally to angular velocities.