自主式水面載具使用軌跡重規劃即時修正圓形編隊控制軌跡誤差之研究

Abstract

移動載具聲層析需要多載具維持特定的隊形來觀測待測海域，本研究以多載具分布於圓形軌跡上並且維持特定的隊形為目標，提出了一種軌跡重規劃的方法來解決編隊過程中軌跡追蹤產生的超越目標點問題。本文之方法首先由線性轉換將固定的圓形軌跡縮放及平移來符合聲層析的取樣路徑以獲得路徑目標點，接著各載具獨立地追蹤相應的目標點來達到載具間的隊形控制。水面載具因環境干擾以及其非線性動力行為而使得軌跡追蹤過程中，容易因控制誤差而超越原設定之目標點，軌跡重規劃即藉由計算時間增量即時推進目標點，避免目標點被載具所超越。本文使用數值模擬及場域試驗驗證在外力干擾情況下此方法之有效性。

本研究設計了一套多載具系統並於台北市內湖區大湖公園實測。在編隊控制時，由於軌跡重規劃為載具針對各自之軌跡追蹤修正，在隊形上仍需群隊間之協調，以保持正確之隊形，在實作中需經由網路共享路徑交換，以具有最大時間的軌跡重規劃載具為基準，來同步所有載具相應目標點的軌跡時間。實驗結果表明了該算法於真實環境中的有效性。

Moving vehicle tomography (MVT) requires multiple vehicles to maintain a specific formation to observe the ocean. This study aims to develop a strategy for multi-Unmanned Surface Vehicles (USVs) to keep on a circular trajectory. The strategy is a trajectory re-planning method to correct the position errors in trajectory tracking caused by the nonlinear behaviors of the USVs and environmental disturbances. A predefined circular trajectory is usually scaled and translated by a linear transformation to conform to the sampling path of the acoustic tomography for obtaining trajectory points. Then each vehicle independently tracks their corresponding trajectory point to achieve the formation between vehicles. Surface vehicles are prone to unexpected speeds due to environmental disturbances and their inertia. This makes vehicles easy to overtake the target point due to too fast speed. Trajectory re-planning advances the target point by calculating time increments to re-direct the vehicle to avoid overturning to track the target point behind the vehicle. Numerical simulations and field tests both verify the effectiveness under external force disturbances.

A multi-vehicle system was designed and tested in Dahu Park, Neihu District, Taipei City. Since the trajectory re-planning is performed by each individual vehicle, the originally defined formation must be coordinated between vehicles. In the implementation, it is necessary to share the route of the all vehicles through a vehicle network. The trajectory time of the target points between different vehicles are synchronized. The experimental results show the effectiveness of the algorithm in the field environment where the network loss and connection delay frequently take place.