Ubuntu18.04-編譯安裝支持Python運動規劃庫OMPL

簡介

最近需要使用到運動規劃庫，於是選擇使用OMPL。主要原因基於兩個點：一、這個庫是ROS默認支持的庫；二、雖然還有CHOMP和STOMP，對比三者OMPL具有更快更穩定的規劃路徑。可以參考這裏看具體差異。下面介紹如何安裝OMPL，這個庫支持多個品臺，Mac,Windows,Ubuntu等。
這個庫安裝太多坑了，文章最後可以留言獲取我編譯好的鏈接庫，可以直接使用。需要的可以留言或者添加公衆號（追逐雅克比）獲取。

開始安裝

1. 坑1 源碼安裝沒有提及依賴項，安裝文檔相對簡單，安裝門檻較高。

這個庫目前是我安裝過最坑爹的庫暫時沒有之一，大家看看這個安裝官方教程，那是相當簡單，太有迷惑性了。只需要下載一個安裝腳本即可，看着相當簡單，其實等你真正安裝哭死你。

2. 坑2 ubutun apt管理源沒有支持app和Python binding

本以爲可以直接使用apt安裝，然後哪一行小小的字提醒你，如果你想用Python，那這是不行的。

第一步：去官網下載源碼包：目前是1.4.2版本。

下載後看到的目錄如下：

第二步：去官網拷貝安裝腳本

第三步：開始安裝----

如果你這就開始安裝，恭喜你，你馬上會遇到各種各樣的錯誤。哪怕最後編譯成功，也可以看到很多依賴庫沒有編譯完成，尤其是python-bindings這個東東。

第四步：安裝依賴

必須要安裝依賴：PyQt5 OpenGL numpy pyplusplus pygccxml flask celery fcl libspot ccd boost eigen3 flann pkg-config
可選依賴：
 * pypy, <http://pypy.org>
   Used to speed up the generation of python bindings.
 * Triangle, <http://www.cs.cmu.edu/~quake/triangle.html>
   Used to create triangular decompositions of polygonal 2D environments.
 * MORSE, <https://www.openrobots.org/wiki/morse>
   OMPL includes a plugin for the MORSE Robot Simulation engine.
 * Drawstuff, <http://ode.org>
   Part of the ODE source distribution, used in one demo program.
 * PQP, <http://gamma.cs.unc.edu/SSV/>
   Used as an alternative, additional collision checking library (the default is FCL).
 * Doxygen, <http://doxygen.org>
   Used to create the OMPL documentation (i.e., http://ompl.kavrakilab.org).

上述依賴有的可以使用apt-get 安裝，有的使用pip安裝，有的安裝完前一個，後面的即可自動安裝：

1. 安裝libspot ，可以參考這裏；
   wget -q -O - https://www.lrde.epita.fr/repo/debian.gpg | sudo apt-key add - echo 'deb http://www.lrde.epita.fr/repo/debian/ stable/' >> /etc/apt/sources.list apt-get update apt-get install spot libspot-dev spot-doc python3-spot(這個庫可以不裝，我沒安裝也成功了，需要Python版本3.7以上) # Or a subset of those
   其他的都可以使用apt-get安裝就不一一寫在這裏了

第五步：真正開始安裝：

在那之前使用如下命令檢查一下配置有沒有遺漏：

cmake -DOMPL_BUILD_PYBINDINGS=ON . && make update_util_bindings VERBOSE=1

如果沒有遺漏，在經過一波操作後會出現如下圖所示：

然後接下來就是漫長的等待，官方告知安裝這個庫編譯需要5-8個小時，可以出去浪一下，讓你們感受一下絕望，連庫自己都擔心你等太久，把程序關了：

第六步：編譯完畢後，檢查這個路徑下ompl-1.4.2-Source/build/Release/lib，是否有如下文件。

注意：兩個紅色箭頭是後面使用python版本的關鍵。如果Python-bindings沒有正確生成Python調用的依賴的話，你後面會出現兩個致命錯誤，爲了解決這兩個錯誤，查看了各種論壇和文檔，才解決，不容易啊。

第一個：
No module named _util
第二個：
No module named _base

第七步：由於需要直接使用我們還希望把動態鏈接庫拷貝到方便調用的地方。我們運行make install。

第八步：去ompl-1.4.2-Source/build/Release/bin下面先測試是否C++ 版本的庫已編譯好。如下圖我們選擇demo_RigidBodyPlanning 進行測試：

第九步：出現已下結果證明已編譯成功

OMPL version: 1.4.2
Info:    RRTConnect: Space information setup was not yet called. Calling now.
Debug:   RRTConnect: Planner range detected to be 1.006980
Settings for the state space 'SE3CompoundSpace0'
  - state validity check resolution: 1%
  - valid segment count factor: 1
  - state space:
Compound state space 'SE3CompoundSpace0' of dimension 6 (locked) [
Real vector state space 'RealVectorSpace1' of dimension 3 with bounds: 
  - min: -1 -1 -1 
  - max: 1 1 1 
 of weight 1
SO(3) state space 'SO3Space2' (represented using quaternions)
 of weight 1
]
Registered projections:
  - <default>
Projection of dimension 3
Cell sizes (computed defaults): [0.1 0.1 0.1]

Declared parameters:
longest_valid_segment_fraction = 0.01
projection.cellsize.0 = 0.1
projection.cellsize.1 = 0.1
projection.cellsize.2 = 0.1
projection.cellsize_factor = 0
valid_segment_count_factor = 1
Valid state sampler named uniform with parameters:
nr_attempts = 100
Start states:
Compound state [
RealVectorState [0.180145 -0.00048426 0.686778]
SO3State [0.706146 0.432641 -0.505694 -0.241771]
]
Goal state, threshold = 2.22045e-16, memory address = 0x55d917523fd0, state = 
Compound state [
RealVectorState [-0.564755 0.867481 0.0330055]
SO3State [0.364372 0.697109 -0.366705 0.496789]
]
OptimizationObjective = nullptr
There are 0 solutions
Info:    RRTConnect: Starting planning with 1 states already in datastructure
Info:    RRTConnect: Created 6 states (2 start + 4 goal)
Found solution:
Geometric path with 5 states
Compound state [
RealVectorState [0.180145 -0.00048426 0.686778]
SO3State [0.706146 0.432641 -0.505694 -0.241771]
]
Compound stadishibute [
RealVectorState [0.430851 0.370329 0.617361]
SO3State [0.434015 0.688351 -0.231927 -0.53293]
]
Compound state [
RealVectorState [0.287885 0.441718 0.533449]
SO3State [0.457808 0.747085 -0.274936 -0.395836]
]
Compound state [
RealVectorState [-0.138435 0.654599 0.283227]
SO3State [0.460797 0.809409 -0.359613 0.0565803]
]
Compound state [
RealVectorState [-0.564755 0.867481 0.0330055]
SO3State [0.364372 0.697109 -0.366705 0.496789]
]



Info:    No planner specified. Using default.
Info:    LBKPIECE1: Attempting to use default projection.
Debug:   LBKPIECE1: Planner range detected to be 1.006980
Properties of the state space 'SE3CompoundSpace3'
  - signature: 6 5 6 1 3 3 3 
  - dimension: 6
  - extent: 5.0349
  - sanity checks for state space passed
  - probability of valid states: 1
  - average length of a valid motion: 2.51125
  - average number of samples drawn per second: sampleUniform()=4.55243e+06 sampleUniformNear()=2.65e+06 sampleGaussian()=2.73573e+06
Settings for the state space 'SE3CompoundSpace3'
  - state validity check resolution: 1%
  - valid segment count factor: 1
  - state space:
Compound state space 'SE3CompoundSpace3' of dimension 6 (locked) [
Real vector state space 'RealVectorSpace4' of dimension 3 with bounds: 
  - min: -1 -1 -1 
  - max: 1 1 1 
 of weight 1
SO(3) state space 'SO3Space5' (represented using quaternions)
 of weight 1
]
Registered projections:
  - <default>
Projection of dimension 3
Cell sizes (computed defaults): [0.1 0.1 0.1]

Declared parameters:
longest_valid_segment_fraction = 0.01
projection.cellsize.0 = 0.1
projection.cellsize.1 = 0.1
projection.cellsize.2 = 0.1
projection.cellsize_factor = 0
valid_segment_count_factor = 1
Valid state sampler named uniform with parameters:
nr_attempts = 100
Planner LBKPIECE1 specs:
Multithreaded:                 No
Reports approximate solutions: No
Can optimize solutions:        No
Aware of the following parameters: border_fraction min_valid_path_fraction range
Declared parameters for planner LBKPIECE1:
border_fraction = 0.9
min_valid_path_fraction = 0.5
range = 1.00698
Start states:
Compound state [
RealVectorState [-0.0730456 -0.836679 0.0699345]
SO3State [0.473221 -0.507003 0.0264253 0.719939]
]
Goal state, threshold = 2.22045e-16, memory address = 0x55d917523fd0, state = 
Compound state [
RealVectorState [0.35548 -0.13107 -0.521166]
SO3State [-0.198537 -0.608689 0.0027985 0.768162]
]
OptimizationObjective = nullptr
There are 0 solutions
Info:    LBKPIECE1: Starting planning with 1 states already in datastructure
Info:    LBKPIECE1: Created 187 (93 start + 94 goal) states in 185 cells (92 start (92 on boundary) + 93 goal (93 on boundary))
Info:    Solution found in 0.000627 seconds
Found solution:
Info:    SimpleSetup: Path simplification took 0.002024 seconds and changed from 133 to 2 states
Geometric path with 2 states
Compound state [
RealVectorState [-0.0730456 -0.836679 0.0699345]
SO3State [0.473221 -0.507003 0.0264253 0.719939]
]
Compound state [
RealVectorState [0.35548 -0.13107 -0.521166]
SO3State [-0.198537 -0.608689 0.0027985 0.768162]
]

第十步：去路徑：omplapp-1.4.2-Source/ompl-1.4.2-Source/demos下找到RigidBodyPlanning.py,使用如下命令進行測試：

 python RigidBodyPlanning.py
 出現以下證明安裝成功：
 Info:    No planner specified. Using default.
Info:    LBKPIECE1: Attempting to use default projection.
Debug:   LBKPIECE1: Planner range detected to be 0.879845
Info:    LBKPIECE1: Starting planning with 1 states already in datastructure
Info:    LBKPIECE1: Created 35 (22 start + 13 goal) states in 32 cells (20 start (20 on boundary) + 12 goal (12 on boundary))
Info:    Solution found in 0.000430 seconds
Info:    SimpleSetup: Path simplification took 0.001344 seconds and changed from 13 to 2 states
Geometric path with 2 states
Compound state [
RealVectorState [0.5 0.521332]
SO2State [0.0343937]
]
Compound state [
RealVectorState [-0.5 0.00545857]
SO2State [-1.72727]
]



Info:    RRTConnect: Space information setup was not yet called. Calling now.
Debug:   RRTConnect: Planner range detected to be 0.879845
Settings for the state space 'SE2CompoundSpace3'
  - state validity check resolution: 1%
  - valid segment count factor: 1
  - state space:
Compound state space 'SE2CompoundSpace3' of dimension 3 (locked) [
Real vector state space 'RealVectorSpace4' of dimension 2 with bounds: 
  - min: -1 -1 
  - max: 1 1 
 of weight 1
SO2 state space 'SO2Space5'
 of weight 0.5
]
Registered projections:
  - <default>
Projection of dimension 2
Cell sizes (computed defaults): [0.1 0.1]

Declared parameters:
longest_valid_segment_fraction = 0.01
projection.cellsize.0 = 0.1
projection.cellsize.1 = 0.1
projection.cellsize_factor = 0
valid_segment_count_factor = 1
Valid state sampler named uniform with parameters:
nr_attempts = 100

Start states:
Compound state [
RealVectorState [-0.690274 0.216669]
SO2State [-1.3863]
]
Goal state, threshold = 2.22045e-16, memory address = 0x55981fa6d9c0, state = 
Compound state [
RealVectorState [-0.98753 -0.931353]
SO2State [-2.71132]
]
OptimizationObjective = nullptr
There are 0 solutions

Info:    RRTConnect: Starting planning with 1 states already in datastructure
Info:    RRTConnect: Created 6 states (2 start + 4 goal)
Found solution:
Geometric path with 5 states
Compound state [
RealVectorState [-0.690274 0.216669]
SO2State [-1.3863]
]
Compound state [
RealVectorState [-0.821073 0.0697578]
SO2State [-0.0200162]
]
Compound state [
RealVectorState [-0.863441 -0.185053]
SO2State [-0.705029]
]
Compound state [
RealVectorState [-0.925485 -0.558203]
SO2State [-1.70817]
]
Compound state [
RealVectorState [-0.98753 -0.931353]
SO2State [-2.71132]
]

• 注意這裏 有可能出現找不到解的情況，這是正常的，多嘗試幾次即可。

最後

由於有的小夥伴實在不想等那絕望的幾個小時，又想直接使用Python版本的ompl可以私信或者留言找我獲取編譯好的安裝包，只需要解壓到Python依賴庫即可。如果有什麼問題也歡迎大家留言，我有時間看到會進行回覆。

參考鏈接

[1] https://bitbucket.org/ompl/ompl/issues/488/no-module-named-_util
[2] https://www.twblogs.net/a/5b83659c2b71776c51e2d839
[3] https://spot.lrde.epita.fr/install.html
[4] http://ubuntuhandbook.org/index.php/2019/02/install-python-3-7-ubuntu-18-04/
[5] https://sourceforge.net/p/ompl/mailman/message/34837507/