无人水面艇自主导航非线性模型预测控制

提升无人水面艇（USV）在复杂海洋环境下的智能水平和作业可靠性，实现高精度轨迹跟踪与安全避碰的协同控制，是当前USV自主导航研究的关键挑战。为此，本文提出一种将国际海上避碰规则（COLREGs）嵌入优化目标的实时非线性模型预测控制（NMPC）框架。通过将COLREGs第14条对遇规则转化为软约束，并结合规则导向的代价函数，构建了轨迹跟踪、避碰与控制平滑性相统一的多目标优化问题，从结构上避免了传统分层控制中的目标冲突，实现全局协调优化。为了满足USV实时性需求，提出了基于动态安全距离的约束集在线压缩方法，通过风险评估动态激活COLREGs约束以降低计算负荷。基于Otter USV的仿真实验表明：控制器可自主执行符合规则要求的右转避让，最小避碰距离为3.44 m，平均计算时间低于20 ms；湖试实验中最小避碰距离为3.14 m。两类实验中的最小避碰距离均大于安全阈值3 m，且控制器在不同场景下均运行稳定。上述结果验证了该框架在安全性与实时性方面的有效性，具有在实际海洋工程中推广应用的潜力。

Enhancing the intelligence and operational reliability of unmanned surface vehicles (USVs) in complex marine environments, and achieving collaborative control for high-precision trajectory tracking and safe collision avoidance, represent key challenges in current USV autonomous navigation research. To address this, this paper proposes a real-time nonlinear model predictive control (NMPC) framework that integrates the International regulations for preventing collisions at sea (COLREGs) into the optimization objective. By converting Rule 14 (head-on situation) of the COLREGs into soft constraints and incorporating a rule-guided cost function, a multi-objective optimization problem is formulated to unify trajectory tracking, collision avoidance, and control smoothness. This integrated structure avoids the goal conflicts inherent in traditional hierarchical control and achieves globally coordinated optimization. To meet the real-time requirements of the USV, an online constraint set reduction method based on dynamic safety distance is proposed, which dynamically activates the COLREGs constraints through risk assessment to reduce computational load. Simulation results based on the Otter USV show that the controller can autonomously execute compliant starboard-turn avoidance maneuvers, with a minimum collision avoidance distance of 3.44 m and an average computation time below 20 ms. Lake trials further confirm a minimum avoidance distance of 3.14 m. Both experimental sets demonstrated a minimum collision avoidance distance greater than the 3-meter safety threshold, along with stable controller operation in all scenarios. The obtained results demonstrate the effectiveness of the proposed framework in ensuring safety and real time performance, highlighting its potential for practical application in ocean engineering.