ASTRA, insansız hava araçlarının karmaşık ve engellerle çevrili ortamlarda güvenli, hızlı ve etkin biçimde hareket edebilmesi için geliştirilen akıllı yörünge planlama projesidir. Projede A-Star algoritması, Ardışık Konveks Programlama (SCP) ve Transformer tabanlı öğrenme yöntemleri bir araya getirilerek çarpışmasız, düzgün ve dinamik açıdan uygulanabilir yörüngelerin düşük hesaplama maliyetiyle oluşturulması hedeflenmektedir.
A-Star Driven Sequential Convex Programming for UAV Trajectory Planning
Recent advances in Unmanned Aerial Vehicles (UAVs) have intensified the demand for autonomous navigation in environments with dense obstacles and varying time dynamics. While traditional trajectory-planning methods based on discrete search or heuristics can efficiently produce collision-free paths or dynamically feasible trajectories, they often struggle to deliver both properties in real-time under non-convex obstacle-avoidance constraints. This paper presents a real-time planning framework based on A-Star and Sequential Convex Programming (SCP) methods, which converts non-convex constraints into convex ones. An A-Star search is used to generate an obstacle-avoiding initial path for reliable initialization, while SCP refines the path into a smooth trajectory. The numerical results show that the proposed A-Star driven SCP approach computes smooth, dynamically feasible trajectories substantially faster than conventional nonlinear solvers, helping to achieve more efficient and responsive UAV operation.
Bu sayfa Marmara Optimization, Signal Processing, Autonomy, and Intelligent Communications Lab tarafından en son 20.11.2026 12:44:03 tarihinde güncellenmiştir.