chakon / parametric_modeling Goto Github PK

1.0 1.0 1.0 4.9 MB

Master Thesis Work on iCub Humanoid Robot (Created Package for Parameteric Modeling)

C++ 100.00%

parametric_modeling's Introduction

The objective is to estimate the ineetial parameters like mass, center of mass & inertia matrx using icub left hand dataset. For more details refer to report of master thesis.

Package includes c++ code for parametric modeling using GURLS package. However first, Install GURLS package with necessary dependencies from https://github.com/LCSL/GURLS. More detail information can be found on GURLS manual. In addition to GURLS you have to install other packages like ICUB, Yarp, kdl_codyco, orocos_kdl, iDynTree. 

iCub & Yarp installation can be found on http://wiki.icub.org/wiki/Linux:Installation_from_sources

kdl_codyco package installation https://github.com/traversaro/kdl_codyco

orocos_kdl package https://github.com/orocos/orocos_kinematics_dynamics/tree/master/orocos_kdl

iDynTree package installation https://github.com/robotology-playground/iDynTree

How to Install Parametric Modeling Package : 

Step 1 : Install all necessary packages & respective dependencies.

Step 2 : root to parametric modeling package version 0.1
	#### cd Parametric_Modeling/Version0.1 ####
Step 3 : make build directory 
	#### mkdir build ####
	#### cd build ####
Step 4 : Cmake
	#### ccmake../ ####
Step 5 : then make it	
	#### make ####
Step 6 : Run the code you want according to your choise for estimating parameters. 

	#### ./iCubParis02_simple_analysis --dataset ../Data_Sets/part1-left.csv --results results.csv #### 

Note : These tests were made on iCub left hand. some tests were made when new package were included for getting regularization parameter. There are certain options to change approach for getting $lambda$. Please go through the code for more details.



This package is simple extension of exisiting work showing the influence of ragularization parameter in estimating inertial parameters. 

Detail explaination of parametric modeling and theory behind this can be found from my thesis which is included in the package. 

Any problems with code or questions can be asked through email [email protected]

parametric_modeling's People

Contributors

Stargazers

Watchers

Forkers

raffaello-camoriano

parametric_modeling's Issues

Faulty results saving

After issuing the command:

./iCubParis02_simple_analysis --dataset ../../Data_Sets/part1-left.csv --results results.csv --xtrain ../../Data_Sets/Xtr_2psmall.txt --ytrain ../../Data_Sets/ytr_2psmall.txt

The program was executed and the results stored in results.csv.

However, the content of results.csv may not be saved properly. This is the file content, which includes several NaNs. How should it be interpreted?

-nan,-nan,-nan,-nan,-nan,-nan,14.3552,-3.7024,-2.86709,0.223453,1.02526,-0.205156,-nan,-nan,-nan,-nan,-nan,-nan,14.3652,-3.66704,-2.86271,0.221296,1.02564,-0.203341,-nan,-nan,-nan,-nan,-nan,-nan,14.3861,-3.58752,-2.85829,0.216157,1.02557,-0.199272,-nan,-nan,-nan,-nan,-nan,-nan,14.4177,-3.48059,-2.83147,0.209381,1.02497,-0.193797,-nan,-nan,-nan,-nan,-nan,-nan,14.4694,-3.2802,-2.80688,0.19641,1.02344,-0.18354,-nan,-nan,-nan,-nan,-nan,-nan,14.4893,-3.21236,-2.78233,0.192061,1.02225,-0.180068,-nan,-nan,-nan,-nan,-nan,-nan,14.5535,-2.96845,-2.71763,0.176628,1.01936,-0.167566,-nan,-nan,-nan,-nan,-nan,-nan,14.5767,-2.87325,-2.69536,0.170671,1.01847,-0.162681,-nan,-nan,-nan,-nan,-nan,-nan,14.6392,-2.62884,-2.60436,0.155183,1.01292,-0.150147,-nan,-nan,-nan,-nan,-nan,-nan,14.705,-2.329,-2.51795,0.136538,1.00776,-0.134744,-nan,-nan,-nan,-nan,-nan,-nan,14.7384,-2.16127,-2.47174,0.126229,1.00503,-0.12612,-nan,-nan,-nan,-nan,-nan,-nan,14.7947,-1.83051,-2.40164,0.106172,1.00125,-0.109095,-nan,-nan,-nan,-nan,-nan,-nan,14.8497,-1.46336,-2.31202,0.0841438,0.996355,-0.0901846,-nan,-nan,-nan,-nan,-nan,-nan,14.8935,-1.11957,-2.22089,0.0636656,0.990688,-0.072467,-nan,-nan,-nan,-nan,-nan,-nan,14.9275,-0.795893,-2.13034,0.0444654,0.983272,-0.0557769,-nan,-nan,-nan,-nan,-nan,-nan,14.6074,-0.605329,-
[FILE IS TOO LONG --> TRUNCATED]

Wrong error measure used in test_for_lambda

hoperf is set to macroavg in test_for_lambda, which is not suitable for regression problems.

See program output:

kammo@kammo-Latitude-E5540:~/ICUB/Parametric_Modelling/Version0.1/build2$ ./test_for_lambda ../../Data_Sets/Xtr_2psmall.txt ../../Data_Sets/ytr_2psmall.txt
=========== Number of rows for training ============

rows of training :1405
============Number of rows displaye d ==============

Specified Task Sequence :
Sequence of size:2

# New task sequence...

[Task 0: split]: ho...  done.
[Task 1: paramsel]: loocvprimal...  done.

Save cycle...
[Task 0: split]: ho... saving
[Task 1: paramsel]: loocvprimal... saving

Saving opt in test_for_lambda.bin
lambda values:
[ 0.000143623 0.000143623 297.176 0.000143623 0.000143623 0.000143623 ]

final lambda value:0.000143623
================= Printing All ================

    [ Name ] = test_for_lambda
    [ calibfile ] = foo
    [ combineclasses ] = Pointer to the function <mean> whose signature is: T (_func)(T_, int)
    [ eig_percentage ] = 5
    [ epochs ] = 4
    _[ hoperf ] = macroavg_
    [ hoproportion ] = 0.2
    [ name ] = test_for_lambda
    [ nholdouts ] = 1
    [ nlambda ] = 20
    [ nsigma ] = 25
    [ paramsel ] = 
~~~~~~~ GurlsOptionList: paramsel
    [ Name ] = paramsel
    [ guesses ] = 
[ 0.000143623 0.000378683 0.000998456 0.00263258 0.0069412 0.0183015 0.0482547 0.127231 0.335463 0.884499 2.33212 6.14897 16.2127 42.7472 112.71 297.176 783.548 2065.94 5447.17 14362.3 ]

```
[ lambdas ] = 
```

[ 0.000143623 0.000143623 297.176 0.000143623 0.000143623 0.000143623 ]

```
[ perf ] = 
```

[ 0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.504274 0 0.992236 0 0 0
  0.495726 0 0.993012 0 0 0
  0.487179 0 0.994565 0 0 0
  0.410256 0 0.995342 0 0 0
  0.34188 0 0.999224 0 0 0
  0.128205 0 1 0 0 0
  0.00854701 0 1 0 0 0
  0 0 1 0 0 0
  0 0 1 0 0 0
  0 0 1 0 0 0 ]

```
    [ perfeval ] = acc
    [ ploteval ] = acc
    [ plotstr ] = test_for_lambda
    [ predbagmethod ] = vote
    [ processes ] = 
~~~~~~~ GurlsOptionList: processes
    [ Name ] = processes
    [ one ] = ( ComputeNsave, ComputeNsave )
```

```
[ randfeats ] = 
```

~~~~~~~ GurlsOptionList: randfeats
    [ D ] = 500
    [ Name ] = randfeats
    [ samplesize ] = 100

```
    [ saveanalysis ] = 1
    [ savefile ] = test_for_lambda.bin
    [ savekernel ] = 1
    [ seq ] = Sequence of size:2
    [ singlelambda ] = Pointer to the function <median> whose signature is: T (*func)(T*, int)
    [ smallnumber ] = 1e-08
    [ split ] = 
~~~~~~~ GurlsOptionList: split
    [ Name ] = split
    [ indices ] = 
gMat2D: (1405 x 1) matrix of type m

    [ lasts ] = 
[ 1124 ]

```

```
[ subsize ] = 50
[ time ] = 
```

~~~~~~~ GurlsOptionList: elapsedtime
    [ Name ] = elapsedtime
    [ one ] = 
[ 0 0.015 ]

```
    [ tmpdir ] = test_for_lambda.bin
    [ todisk ] = 1
    [ verbose ] = 1
    [ version ] = 2.0
~~~~~~~ ================= Printed All ================
```

Wrong error measure used in iCubParis02_simple_analysis

Same problem as in Issue #3. The selected lambda is not reliable.

kammo@kammo-Latitude-E5540:~/ICUB/Parametric_Modelling/Version0.1/build2$ ./iCubParis02_simple_analysis --dataset ../../Data_Sets/part1-left.csv --results results.csv --xtrain ../../Data_Sets/Xtr_2psmall.txt --ytrain ../../Data_Sets/ytr_2psmall.txt
1405

Specified Task Sequence :
Sequence of size:2

# New task sequence...

[Task 0: split]: ho...  done.
[Task 1: paramsel]: hoprimal...  done.

Save cycle...
[Task 0: split]: ho... saving
[Task 1: paramsel]: hoprimal... saving

Saving opt in iCubParis02_simple_analysis.bin
lambda values:
[ 0.00014291 0.00014291 112.15 0.00014291 0.00014291 0.00014291 ]

final lambda value: 0.00014291
================= Printing All ================

    [ Name ] = iCubParis02_simple_analysis
    [ calibfile ] = foo
    [ combineclasses ] = Pointer to the function <mean> whose signature is: T (_func)(T_, int)
    [ eig_percentage ] = 5
    [ epochs ] = 4
    [ hoperf ] = macroavg
    [ hoproportion ] = 0.2
    [ name ] = iCubParis02_simple_analysis
    [ nholdouts ] = 1
    [ nlambda ] = 20
    [ nsigma ] = 25
    [ paramsel ] = 
~~~~~~~ GurlsOptionList: paramsel
    [ Name ] = paramsel
    [ guesses ] = 
[ 0.00014291 0.000376804 0.000993502 0.00261952 0.00690675 0.0182107 0.0480153 0.126599 0.333799 0.88011 2.32054 6.11846 16.1323 42.5351 112.15 295.701 779.66 2055.69 5420.14 14291 ]

```
[ lambdas ] = 
```

[ 0.00014291 0.00014291 112.15 0.00014291 0.00014291 0.00014291 ]

```
[ lambdas_round ] = 
```

[ 0.00014291 0.00014291 112.15 0.00014291 0.00014291 0.00014291 ]

```
[ perf ] = 
```

[ 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.428571 0.357143 0.107143 0.0357143 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 0.996047 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ]

```
    [ perfeval ] = acc
    [ ploteval ] = acc
    [ plotstr ] = iCubParis02_simple_analysis
    [ predbagmethod ] = vote
    [ processes ] = 
~~~~~~~ GurlsOptionList: processes
    [ Name ] = processes
    [ one ] = ( ComputeNsave, ComputeNsave )
```

```
[ randfeats ] = 
```

~~~~~~~ GurlsOptionList: randfeats
    [ D ] = 500
    [ Name ] = randfeats
    [ samplesize ] = 100

```
    [ saveanalysis ] = 1
    [ savefile ] = iCubParis02_simple_analysis.bin
    [ savekernel ] = 1
    [ seq ] = Sequence of size:2
    [ singlelambda ] = Pointer to the function <median> whose signature is: T (*func)(T*, int)
    [ smallnumber ] = 1e-08
    [ split ] = 
~~~~~~~ GurlsOptionList: split
    [ Name ] = split
    [ indices ] = 
gMat2D: (1405 x 1) matrix of type m

    [ lasts ] = 
[ 1124 ]

```

```
[ subsize ] = 50
[ time ] = 
```

~~~~~~~ GurlsOptionList: elapsedtime
    [ Name ] = elapsedtime
    [ one ] = 
[ 0 0.001 ]

```
    [ tmpdir ] = iCubParis02_simple_analysis.bin
    [ todisk ] = 1
    [ verbose ] = 1
    [ version ] = 2.0
~~~~~~~ ================= Printed All ================
```

Linking issue

-- Found YCM: /home/kammo/Repos/codyco-superbuild/build/install/share/YCM (found version "0.2.0~20141017+gita2285614")
Looking for KDL in: /home/kammo/Repos/codyco-superbuild/build/install
-- Using iCub from install
-- Configuring done
-- Generating done
-- Build files have been written to: /home/kammo/ICUB/Parametric_Modelling/Version0.1/build2
[ 11%] Building CXX object CMakeFiles/Data_Model_trail.dir/Data_Model_trail.cpp.o
/home/kammo/Repos/Parametric_Modeling/Version0.1/Data_Model_trail.cpp:1:0: warning: "NDEBUG" redefined [enabled by default]
#define NDEBUG
^
:0:0: note: this is the location of the previous definition
Linking CXX executable Data_Model_trail
/usr/bin/ld: cannot find -lkdl_codyco
/usr/bin/ld: cannot find -lkdl_format_io
/usr/bin/ld: cannot find -liDynTree
collect2: error: ld returned 1 exit status
make[2]: *** [Data_Model_trail] Error 1
make[1]: *** [CMakeFiles/Data_Model_trail.dir/all] Error 2
make: *** [all] Error 2
*** Failure: Exit code 2 ***

chakon / parametric_modeling Goto Github PK

parametric_modeling's Introduction

parametric_modeling's People

Contributors

Stargazers

Watchers

Forkers

parametric_modeling's Issues

Faulty results saving

Wrong error measure used in test_for_lambda

# New task sequence...

Wrong error measure used in iCubParis02_simple_analysis

# New task sequence...

Linking issue

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent