multistage based parking planner
2022 Dftpav: Flatness-Based Trajectory
- analytic trajectory representation with differential flatness property
- efficient obstacle avoidance with a safe driving corridor for unmodelled obstacles
- signed distance approximations for dynamic moving objects
Ref
- An Efficient Spatial-Temporal Trajectory Planner for Autonomous Vehicles in Unstructured Environments
- Differential Flatness-Based Trajectory Planning for Autonomous Vehicles
- code
Related Work
- trajectory optimization:
- convex elastic band smoothing (CES): eliminates the non-convexity of the curvature constraint with fixed path lengths, transform the original problem into a quadratically constrained quadratic program (QCQP)
- dualloop iterative anchoring path smoothing (DL-IAPS): sequential convex optimization (SCP) is used to relax the curvature constraint
- obstacle avoidance
- two circle to cover ego car
- OBCA: 增加优化问题维数; 对偶元和障碍物数量成正比
- H-OBCA, with MPC
SPATIAL-TEMPORAL TRAJECTORY PLANNING
adopt the lightweight hybridA* algorithm to find a collision-free path that is further optimized by the proposed planner.
differential flatness
1995: Differential flatness of mechanical control systems: A catalog of prototype systems
对非线性系统通过有限的微分近似进行降维. 对于机器人的运动规划,可以将所有的轨迹约束映射到平坦输出空间,降低优化维数.得到最优解之后再上升回到初始的状态空间中.
比如对于二维车辆,其独立变量理论上只需要两个量.假设取平坦输出为$x, y$, 输入取速度和前轮转角.则可以得到关系式:
此时规划轨迹可以只需要得到$x(t), y(t)$. 一般可以选择基函数$\beta(t)$组合成平坦输出:
比如常见的把$x,y$表示为多项式进行表示.
Differentially Flat Vehicle Model:
"with the natural differential flatness property,we can use the flat outputs and their finite derivatives to characterize arbitrary state quantities of the vehicle, which simplifies the trajectory planning and facilitates optimization".
avoid singularities by fixing the velocity magnitude to a small, non-zero constant when the gear shifts. Both the gear shifting position and directional angle can be optimized(优化换挡点).
INSTANTANEOUS STATE CONSTRAINTS
Dynamic Feasibility
Static Obstacle Avoidance
generate free convex polygon based on the sampling point(2020 Generating large convex polytopes directly on point clouds, code).
H-representation: On canonical representations of convex polyhedra
2022 FTHA && Iterative NLP
- 通过hybrid A* 寻找相对较优的轨迹, 避免搜索失败
- 圆形代替车身; 基于圆形的二维C空间; 构建方形corridor
- 罚函数代替边界约束条件; 多次求解, 根据当前最优轨迹重新构建corridor
ref
- Optimization-based trajectory planning for autonomous parking with irregularly placed obstacles: A lightweight iterative framework
- github
- bilibili
Abstract
- focused on optimization base parking
- within-corridor collision constraints: scale is small and independent from environment complexity
- reconstructs the corridor in an iterative framework
Intro
- samping / seach based
- optimization based
contributions:
- a fault-tolerant variant of the hybrid A∗ algorithm is proposed
- an iterative framework that incorporates a corridor construction procedure and a lightweight NLP solution procedure
Problem formulation
bicycle kinematic model:
为了避免$\theta \in [0, \pi]$的不连续,边界条件以三角函数约束代替值约束:
Collision avoidance
根据三角形面积构造不等式
Corridor Construction
对于给定的初始路径,可以只考虑该初始值的"附近空间结构",而不需要考虑环境中所有的障碍物,这样就避免过多的障碍物导致的复杂计算.
Generation of a Guiding Route
fault-tolerant hybrid A∗ (FTHA):
- 如果算法失败,就选择当前为止最优的结点,通过A*与终点连接
- 添加简单的速度信息生成初解
construction of corridors
沿着四个方向按固定步长逐渐增长,若失败或者达到最大长度则停止扩大
Within-Corridor Constraint Formulation
障碍物约束改写为圆在corridor
里面:
$P_{\mathrm{f}}(t)$ stays inkth local box for $P_{\mathrm{f}}.$
$P_{\mathrm{r}}(t)$ stays inkth local box for $P_{\mathrm{r}},$, $t = T/N_{FE}·k,k =1,\cdots,N_{FE}$
LIGHTWEIGHT ITERATIVE OPTIMIZATION FRAMEWORK
penalty function to merge boundary constraints
iteratively solve the ocp problem and generate corridors based on the optimal trajectory util the infeasibility is reduced to tiny
2021 tiny passage
Ref
freespace规划
- A* 发现窄通道
- 混合A* 连接窄通道
- 基于优化方案进行后处理
后处理的碰撞检测使用within-STC constraints
TDR-OBCA
H-OBCA
Ref
- An Efficient Spatial-Temporal Trajectory Planner for Autonomous Vehicles in Unstructured Environments
- Differential Flatness-Based Trajectory Planning for Autonomous Vehicles
- 2015 CES: A convex optimization approach to smooth trajectories for motion planning with car-like robots
- 2021 Autonomous driving trajectory optimization with dual-loop iterative anchoring path smoothing and piecewise-jerk speed optimization