Inverse Kineamtics Learning Model for 2 Link Manipulator

ANN model fitted on a data from a 2 link manipulator forward kineamtics to drive the inverse kinematics of the manipulator and test it on a simulated and visualized model.

Install

Dependencies

• numpy
• pandas
• matplotlib
• livelossplot
• scikit-learn
• tensorflow

git clone https://github.com/AbdallahAmrBeedo/Inverse_kinematics_learning_model.git

cd Inverse_kinematics_learning_model/

chmod a+x install_dep.sh

./install_dep.sh

Tutorial

Dataset Generation

to generate the dataset for the manipulator, run

python3 gen_data_set.py

A csv will be created in /data directory with 6400 data points.

	x	y	theta	q1	q2
0	2.0000	0.0000	0.000000	0.000000	0.000000
1	1.9974	0.0715	0.071581	0.000000	0.071581
2	1.9898	0.1427	0.143161	0.000000	0.143161
3	1.9770	0.2131	0.214742	0.000000	0.214742
4	1.9593	0.2824	0.286322	0.000000	0.286322
...	...	...	...	...	...
6395	0.8369	-1.5874	11.023411	5.654867	5.368544
6396	0.9083	-1.5828	11.094992	5.654867	5.440125
6397	0.9792	-1.5732	11.166572	5.654867	5.511706
6398	1.0492	-1.5585	11.238153	5.654867	5.583286
6399	1.1180	-1.5388	11.309734	5.654867	5.654867

6400 rows × 5 columns

Trainning ANN Model

A model with 6 Hidden layers and 10272 trainable paramters in model_train.ipynb just run the cells to train, evaluate and save the model.

Visualization

run

python3 inv_kinematic_viz.py

wait for the model to be loaded, enter x, y, theta you desire and the model will solve the inverse kinematics and plot the manipulator to verify.