This work presents the implementation of a biologically inspired and adaptive robot control system on neuromorphic hardware. The control system is built from spiking neurons organised in a structure inspired by the cerebellum. The network learns to perform the vestibulo-ocular reflex (VOR) in real-time. The system is shown to perform the task in simulated and physical environments.

This work uses a model designed by Naveros et al (F. Naveros, J. A. Garrido, A. Arleo, E. Ros and N. R. Luque, Exploring Vestibulo-Ocular Adaptation in a Closed-Loop Neuro-Robotic Experiment Using STDP. A Simulation Study, 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, 2018, pp. 1-9, doi: 10.1109/IROS.2018.8594019.).

The model is ported onto SpiNNaker and simulated using sPyNNaker. The circuit's ability is shown in a tightly-coupled simulated environment (SpiNNGym), in a loosely-coupled simulated environment (NRP) and in a physical robot (iCub).

This installation assumes the existence of either sPyNNaker or NEST, for brevity. For installation instructions for NEST, please follow official directions. For SpiNNaker, 3 shell scripts are provided for quick installation.

A complete installation (assuming no errors) could be done by running (in this order):

./install.sh && ./setup_develop.sh && ./automatic_make.sh

The software package can be installed in a local virtual environment. The following can be run inside the unarchived folder to install the package:

python setup.py develop

or

pip install .

Each experiment script is stand-alone. However, full-scale experiments can have arguments passed to them via the argparser.

When citing this work, the authors would prefer you to cite the related journal article.