`biotac_sim_plugin`

A metapackage for simulating the tactile data from BioTac sensors mounted on a Shadow Dexterous Hand.

Quick Start

To get started, simply clone the necessary dependency and

Clone this repository using the command below

git clone git@github.com:vmstavens/biotac_sim_plugin.git

Source your catkin workspace using

source <catkin-workspace>/devel/setup.bash

Build the package using
```
catkin build biotac_sim_plugin
```

Run the provided demo

roslaunch biotac_sim_plugin biotac_sim_demo.launch

Packages

The project consists of two packages: biotac_sim_lib and biotac_sim_demo.

`biotac_sim_lib`

The biotac_sim_lib package is responsible for generating the Gazebo model plugin for simulating the BioTac tactile sensor data. The package include the following header files

biotac_sim_lib.hpp which is the header file containing the overall plugin structure.
neural_network.hpp which contains the NeuralNetwork class which is responsible for the deep learning part including loading and running the model.
helpers.hpp which contains the necessary template functions, classes and structs used in neural_network.hpp and biotac_sim_lib.hpp.

Furthermore, the package contains config/model.yaml which is the model architecture and trained parameters, scripts/set_default_model which sets the default model to the one mentioned previously and finally src/* which contains the source files for the above mentioned headers.

This package is not meant to be run directly and therefore contains no launch/example files.

`biotac_sim_demo`

The biotac_sim_demo package is responsible for running a demo of the biotac_sim_lib. This can only be done in the Shadow Dexterous docker container environment, which can be installed as instructed here.

The demonstration can be run using the following command

roslaunch biotac_sim_plugin biotac_sim_demo.launch

This package contains several folders

examples/ which contains biotac_sim_demo.py, the example node run from the command above.
launch/ which contains biotac_sim_demo.launch, the launch file running the node in biotac_sim_demo.py.
models/ which contains a 3D model of a black pen, used as a prop for the demonstration.
robots/ which contains wrapper launch files and xacro files for loading the plugin from biotac_sim_lib
worlds/ which contains two worlds: one with and one without gravity. The world you want can be set in the launch file or run using the following command
```
roslaunch biotac_sim_plugin biotac_sim_demo.launch gravity:=true
```
or gravity:=false in case you want gravity disabled.

Dependencies

Outside of standard ROS and Gazebo packages, the following dependency is needed

ros_utils, which is a utilities package for ROS used for base64 encoding and decoding.

Acknowledgments

This work is based on the biotac_gazebo_plugin by Philipp Ruppel, Yannick Jonetzko, Michael Görner, Norman Hendrich and Jianwei Zhang - Simulation of the SynTouch BioTac Sensor, The 15th International Conference on Intelligent Autonomous Systems, IAS-15 2018, Baden Baden, Germany.

The data set used to train the deep learning model can be found here

The 3D model used for demonstration is part of the dataset: A. Rasouli, J.K. Tsotsos. "The Effect of Color Space Selection on Detectability and Discriminability of Colored Objects."