Code for ICDL-Epirob 2014
Reusing Motor Commands to Learn Object Interaction
The code corresponding to the ICDL 2014 paper Reusing Motor Commands to Learn Object Interaction is available in two forms.
Easy to Run Examples
Some small examples are available illustrating goal babbling and the reuse algorithm. They run in a few seconds, and allow to understand the effects of differents parameters on the exploration strategy quickly, and are avaible here.
Actual Experimental Code
The experiments described in the paper involve much more code, to orchestrate the The whole experiment involve a much larger set of code, and use, in particular, the V-REP simulator. You can run the simulations without any specific hardware, and a full hardware setup.
Examining the code
If you are interested in examining the code, the most relevant packages are
github.com/humm/explorers that contains the actual exploration algorithms,
github.com/humm/experiments where the experiments are described and that organise how things are run, and
github.com/humm/dovecot that organize the code for the simulation and hardware setups.
Running the code
We do not advise trying to run the code at the moment. We are in the process of making the installation smoother, in particular for If you actually want to re-run the experiment, some work may be involved, and we would be happy to help (contact us at email@example.com). You will need to install V-REP, fcl, dmpbbo and the following packages:
https://github.com/humm/toolbox https://github.com/humm/clusterjobs https://github.com/humm/models https://github.com/humm/powerswitch https://github.com/humm/pyvrep https://github.com/humm/natnet https://github.com/humm/pydyn https://github.com/humm/vworlds https://github.com/humm/dovecot # orchestrating code for the environment https://github.com/humm/experiments # experiments are defined and run there