Real-Time Trajectory Planning for Autonomous Urban Driving: Framework, Algorithms, and Verifications
目的
- generating dynamically feasible trajectories in real time with limited onboard computational resources;
- dealing with the unpredictably changing surrounding environments with limited sensing range and visibility as well as the uncertainty and noise in the perception and localization systems;
3)interacting with other traffic participants, such as vehicles, cyclists, and pedestrians
系统框架
Sensors：感知系统和精确定位
结果如图，粉色为检测到的障碍，白线为地图上的数据
we employ a nonlinear optimization approach to smoothen the centerline, then interpolate it using the cubic B-spline curve.
结果：光滑的线
为什么解耦成横向和纵向误差
As referred in [13], to mimic human-like driving behaviors in urban environments, the vehicle trajectory could be naturally decoupled into thelateralmovementandthelongitudinalmovement.
空间坐标、速度坐标基于曲线坐标系
the trajectory planning taskis decomposed into spatial path planning and velocity planning subtasks based on the curvilinear coordinate instead of the Cartesian coordinate
系统算法
the graph-search algorithm 图搜索算法找出无碰撞的参考路线
A、参考路径光滑和插值
the right boundary and left boundary of the desired lane
直接计算出参考路径，但不连续或者不光滑，不容易底层控制器的控制，
会使控制过载或震荡（ overshoots and oscillations）
急转弯时，会有很大的横向加速度
共轭梯度优化算法（ the conjugate gradient optimization method）
作用：使参考路线光滑
局部最优解：Like the other nonlinear optimization methods, the conjugate gradient algorithm only returns a local optimal solution.
顶点距离有时过大，采用三次B-样条插值
To obtain dense waypoints, we use the cubic B-spline for interpolation.
environmental constraints 环境约束没有考虑
B、 Spatial Path Generation Based on the Curvilinear Coordinate Framework （基于曲线坐标框架的空间路径生成）
the sampling-based motion planning strategy
trajectories that are aligned withthe road geometry
更安全，跟友好，缩小解空间
曲线坐标系分别处理汽车横纵向移动（lateral and longitudinal movements of vehicles）
we decompose the trajectory planning task into spatial path generation （空间路径生成）and velocity profile generation（速度剖面生成） subtasks
经过处理的参考路径作为基准坐标系（the refined reference path is taken as the base frame）
