基於階層式全姿態控制與重組控制分配之過驅動向量推進模組無人機於安全空中運輸應用

Abstract

隨著電商盛行，物流運輸達到新的高峰，最後一哩路的成本與效率變得至關重要。離島與山區等偏遠地區也亟需新的解決方案，除了本身運輸成本高昂的問題，若發生天災如地震、颱風導致聯絡道路失效，更有可能造成財產甚或生命上的重大損失。

本研究提出具有同軸螺旋槳設計之向量控制模組化群組無人機，藉由高冗餘、高可控性、可重組等特性，提升安全性、飛行效率、可操作性、機動性，提升抵抗外界干擾與馬達失效等能力，解決現有無人機對於航程、安全、效率等需求。無人機的模組化設計使其能夠通過改變模組類型與數量來因應不同的貨物大小與重量，以實現最佳任務性能。

本研究針對提出的設計推導其數學模型並探討其動態特性，包含單台無人機系統中的欠驅動系統，與群組無人機於不同組態下產生不同的操作空間的過驅動系統。基於估測不同組態下的可操作力空間，使得控制器可以自動調整控制器參數，當組態改變時，毋須重新設計控制器參數。本研究提出針對單台無人機以及群組無人機的控制器，並採用整合完全姿態控制器與重組式控制分配的階層式控制架構。首先，包含具有姿態規劃的全姿態追蹤控制器負責計算虛擬控制指令，使系統可以追蹤參考軌跡。接著提出精確捆綁重組式控制分配方法(EBRCA)，其中包含一個虛擬邊界保護機制來解決黏壁問題，以及一個力矩增強方法來增強力矩分配，以應對非線性、耦合和受限制的可操作力矩空間。

為驗證提出系統之特性，在模擬中，系統被施加風吹、快速路徑變更、模組失效等干擾，並針對異向與同向組態的可控性討論其對於激進動作、非零姿態的路徑追蹤效果。全姿態控制器和控制分配演算法的有效性於靜態、動態和閉迴路測試的模擬中獲得驗證。模擬結果表明，所提出的機構設計與控制器在高機動性、非零姿態軌跡以及不同干擾條件下，例如未知負載重量、突然軌跡變化和未預期的馬達故障中表現良好。在模擬中亦進行EBRCA與CGI和SQP方法的性能比較，結果顯示EBRCA在力矩分配響應獲得更小誤差，並在軌跡追蹤任務中表現較短的穩定時間。此外亦發現，在不一致方向的群組配置下，無論是在靜態穩定還是動態追蹤任務中，都能實現較小的穩態誤差。在姿態穩定與繫留飛行的實驗中，控制架構與無人機設計的基本有效性已被初步驗證。

In this research, a Thrust-Vectoring Modular Drone (TVMD) with coaxial rotors is proposed to address delivery challenges, such as the last-mile delivery for e-commerce services and transporting time-sensitive packages to outlying islands and mountain areas, especially when natural disasters or accidents occur. The TVMD is designed with high redundancy, controllability, and reconfigurability to ensure safety, flight efficiency, mobility, and maneuverability. The modular design of the drone enables optimal task performance by changing the type and number of agents.

The dynamic model of the TVMD is derived and analyzed, which reveals the under-actuation property of the single-agent system and the over-actuation property of the teamed system. The proposed attainable space approximation algorithm estimates system characteristics in various team configurations, which also makes it possible to automatically calculate proper parameters for different team configurations without the need to re-design the controller. A cascaded control architecture is adopted for the teamed systems. Firstly, a full-pose trajectory tracking controller with an attitude planner takes the response to calculate virtual controls and forces the system to track the reference trajectory. Secondly, an exact bundled redistributed control allocation (EBRCA), incorporating pseudo-boundary protection (PBP) for the wall sticking issue and a post-torque enhancement (PTE) to strengthen torque allocations, is proposed to deal with the nonlinear, coupled, and constrained admissible force space.

The effectiveness of the full-pose tracking controller and the control allocation is evaluated using static, dynamic and closed-loop tests in simulations. Simulation results demonstrate that the TVMD performs well in aggressive maneuvers, non-zero attitude trajectory tracking, and under various disturbances such as unknown payload weights, sudden trajectory change, and unexpected motor failures. The performance of the proposed EBRCA is compared with the cascaded generalized inverse (CGI) and the sequential quadratic programming (SQP) methods, and it shows that EBRCA has better torque allocation responses and results in a shorter settling time in tracking tasks. Also, inconsistent orientation team configurations achieve smaller steady-state errors in both regulation and tracking tasks. Preliminary experiments including attitude stabilization and tethered hovering are conducted to verify the basic effectiveness of the proposed control architecture and the UAV design.