当前位置:   article > 正文

UMVs Motion Planning_综述

motion plan

Introduction

  • Preliminary and Technical Review

  • Optimization-based Trajectory Planning

  • Search-based Trajectory Planning

  • Conclusion and Future Work

Background

  • Various mobile platforms

    image-20201029190614636

    • Why choose MAVS
      • 垂直升降
      • 极限运动(较大的加速度)
      • 体积小,可以在狭小空间运动
    • 非常理想的planning&control 平台
  • MAVs are capable of VTOL and 3D motions

    • Fly in an unstructured environment
    • fly in confined indoor/outdoor environments
  • A fully autonomy MAV

    • 激光雷达:无人机周围障碍物检测,描绘地图
    • RGB相机:找到感兴趣的障碍物或目标点
    • 双目摄像头:状态估计,定位,测速,位姿等
    • laser Beam: 激光, 测量高度 作用等同于气压计
    • Autopilot: IMU等,转换控制信号至发动机信号
    • 机载电脑:computer
  • Control Diagram

image-20201029191837381

  • How to evaluate planner(metrics):

    • Feasibility: 是否可以被MAVs执行;
    • Safety:是否无碰撞;
    • Optimality: 是否最优或者sub-optimal;
    • Completeness: 如果存在结果,planner是否可以找到它,而不会丢失解;
    • Run time:找到结果的运行时间, 动态环境中re-plan;
  • Related work

    • Path planning(只包含空间信息)

      • Deterministic: Dijkstra, A* , D* , D* Lite/LPA * , ARA * (精确式)
      • Randomized: PRM, RRT, RRT *, FMT, BIT * (随机采样)
    • Trajectory planning(包含时空信息)

      • Unconstrainted QP(非限制二次规划)

      • MIP(Mix Integer Programming)(找到空中走廊,计算复杂,比较慢)

    • local planner

      • 计算很快
      • 不能提供全局最优解
      • 有时候会被track在一个死胡同中

image-20201029193557864

  • Conclusion of previous work:

image-20201029194635171

  • 状态表示(微分平坦: 做planning的时候只需要考虑 Flat output):

image-20201029194806934

image-20201029195135834

  • 确定性 vs 随机性

Optimization-based Trajectory Planning

  • Piece-wise Polynomial (多段式的多项式)

image-20201029195756391

image-20201029195957488

  • Unconstrained QP

    • Cost Function in terms of coefficients:

    image-20201029200345390

    • Waypoint Constraints:

      • 连续性(空间、速度、加速度、角速度、角加速度等)

        image-20201029200439582

        image-20201029200453940

        • problem re-formulation

      image-20201029200737973

image-20201029201040943

  • Using QP
    • Solve QP with a Safe Flight Corridor(SFC)

image-20201029201612300

  • Run time analysis

image-20201029201711644

  • Problem
    • Time allocation
    • searching homology class

image-20201029202132477

  • Conclusion

image-20201029202249229

Search-based Trajectory Planning

  • Problem revisit

    image-20201029230557760
image-20201029230652773
  • Motion primitives(运动原语,可执行的最小的移动单元)

    • sampling in state space、

      image-20201029232219666
    • sampling in control space

image-20201029232314746

  • comparing to optimization-based method在这里插入图片描述

  • List of contents:

    • Trajectory planning with motion uncertainty
    • Trajectory planning with limited field-of-view(FOV)
    • Trajectory planning in SE(3)
image-20201029233044978
  • Motion uncertainty

    • Motivations

      • High control authority is impractical

      • Real-world environmental factors : wind, air-drag, wall effect.

      • Over-inflating obstacles is not a complete solution
        在这里插入图片描述

      • 能量场

在这里插入图片描述

  • Limited FOV

    • Challenges in real world navigation

      • Map is unknow
      • sensor has limited FOV
    • Desired yaw is non-linear and coupled:

      image-20201029234543115
    • Requirement for safe navigation:

      • The MAV should always move in the direction that is visible在这里插入图片描述
  • Add yaw cost and enforce a hard yaw constraint

image-20201029234734101

  • SE(3)
    • 传统把无人机当做球体,因此可以忽略row和pitch
      • 将其当成椭球体可以进行更加极限的运动规划在这里插入图片描述

conclusion

image-20201030000438480

声明:本文内容由网友自发贡献,不代表【wpsshop博客】立场,版权归原作者所有,本站不承担相应法律责任。如您发现有侵权的内容,请联系我们。转载请注明出处:https://www.wpsshop.cn/w/喵喵爱编程/article/detail/879177
推荐阅读
相关标签
  

闽ICP备14008679号