基于协作神经动力学优化的四旋翼轨迹规划

Quadrotor Trajectory Planning Based on Collaborative Neurodynamic Optimization

摘要: 针对四旋翼轨迹规划中的非凸约束与局部最优问题，本文提出一种基于协作神经动力学优化的轨迹规划方法。利用微分平坦性将问题转化为样条参数优化，并将动力学约束与飞行走廊统一建模为可微惩罚函数。采用分布式实现多节点协同优化，并引入快速随机跳跃算法增强全局搜索能力。结果表明，该方法可生成满足速度与加速度约束的平滑轨迹，并有效降低目标函数，验证了方法的有效性与可行性。

Abstract: To address the non-convex constraints and local optima problems in quadcopter trajectory planning, this paper proposes a trajectory planning method based on collaborative neurodynamic optimization. By utilizing differential flatness, the problem is transformed into a spline parameter optimization problem, and the dynamic constraints and flight corridor are uniformly modeled as differentiable penalty functions. A distributed implementation is used to achieve multi-node collaborative optimization, and a Fast Stochastic Jump algorithm is introduced to enhance global search capabilities. The results demonstrate that this method can generate smooth trajectories that satisfy velocity and acceleration constraints while effectively minimizing the objective function, thereby validating the method’s effectiveness and feasibility.