Benedikt Feldotto, one of the NRP researchers at TUM, has provided a detailed introduction of the NRP architecture and tools to the students of the ACROSS International Summer School on Neurorobotics.

With the presentation "Embodied Learning with the Neurorobotics Platform", Benedikt explained how the Neurorobotics Platform (NRP), a cloud based simulation platform developed in the Human Brain Project, can help users interconnect artificial and spiking neural networks with physics simulations of robots and musculoskeletal bodies. Several example experiments were provided, including an embodied million neuron rodent brain deployed on a High Performance Computing cluster.

Two posters submitted by Fortiss' researchers Michael Neumeier and Evan Eames got accepted and the authors had the opportunity to present tem to a very interested audience during the second day of the event.

The posters with the titles "The Neurorobotics Platform: Neuromorphic solutions both simulated and real" and "Spiking Reinforcement Learning for Robotic Object Insertion on Loihi" showcased results of experiments run in the NRP, the first on a visual place recognition task and the second one on a classic robotic object insertion task, the peg-in-hole. For this specific second experiment, the trained actor network was moved onto neuromorphic hardware (Intel’s Loihi Neuromorphic Chip) and then connected to a real Kuka LBR iiwa robotic arm with a force-torque sensor implemented on the end effector, allowing the neuromorphic hardware to control the movement of the arm, and guide it through the insertion, in real time.

Fortiss' researcher Michael Neumeier presented the latest achievements and features of the NRP at the 2022 Telluride Neuromorphic Cognition Engineering workshop.
His presentation, "The Neurorobotics Platform: Neuromorphic solutions both simulated and real" Michael included the following highlights:

• The Neurorobotics Platform with its architecture and GUI
• Outlook on NRP 4.0
• Neuromorphic solutions in simulation (swimming robot, drone with event camera, robotic arm)
• Ported to real prototype (Kuka robotic arm).

About 25 participants attended Michael's presentation during the LS2 session of the event and had later on the opportunity to join a hands-on exercise on how to set up an NRP experiment where users can create a digital twin in the NRP for a visual place recognition task.

This new release is based on Gazebo 11.3, which means that sensor plugins now output a consistent number of samples at each time step of the physics engine. The Docker-based local installation of the NRP can now use Jupyter notebooks to run the Virtual Coach interactively. And finally, as requested by our users, the popular Intel Loihi neuromorphic chip can now run as a backend for simulating spiking neural networks on the NRP.

But that's not all! V3.2 is going to be the last major release of the legacy NRP and NRP 4.0 will feature a brand new software architecture supporting experiments based on a much expanded range of simulators (e.g. OpenSim, TVB, etc.). Stay tuned!

At this year event, the NRP was one of the chosen platforms available to participants of the SMI21 Sensory Motor Integration topic area. As explained by the workshop organisers, the goal of this specific topic area was to "apply principles of brain-inspired sensori-motor control as observed in biological organisms to artificial agents in a naturalistic simulated environment for solving navigation tasks".

A video of the workshop recording both presentations and activities is available on YouTube here. The SMI Sensory-Motor Integration topic area section starts at 01:48:59.

Our paper, "Massively Parallel Robot Simulations with the HBP Neurorobotics Platform", makes the case of how virtual experiments run in the NRP can replace physical setups and accelerate learning through parallelization. Specifically, we show how after introducing an extension to the NRP we were able to run massively parallel experiments with a full-featured robot simulator and successfully carry out an accelerated collection of synthetic data sets.

Access the paper here for more details, and don't forget to check our poster here!

Dr Fabrice Morin, (NRP), and Dr Paul Triebkorn, (VEP), will present “Brain simulation for clinical applications and neuroscience research: the EBRAINS Virtual Epileptic Patient and Neurorobotics Platform”, introducing main concepts and goals of each technology, how to operate them, and how they can be used in combination to address specific research questions.

Regisration to this exciting two day event open to anyone is only £1.50 and all profits will be donated to ActionDuchenne.org.

A short video illustrating our achievement.

The main goal of the RoboBrain project is to carry out embodied simulation of a large-scale, multi-area brain model running on the NRP and deployed on Japan's new flagship supercomputer Fugaku. Presently, a large-scale whole-brain model (NEST 3.0) is successfully deployed on the NRP and running on the HBP HPC infrastructure (Piz Daint Cluster at CSCS, Lugano, CH).
While our ultimate goal is to recreate in silico a specific in vivo experiment, with this first step we show how to implement an experimental setup on the NRP while providing a direct mapping from brain activity to muscle activation.
Watch our video to learn how we did it!

NRP 3.1 is based on Ubuntu 20.04, and Python 3.8, and users will be able to rely on ROS Noetic instead of ROS Melodic.
NRP 3.0.5. instead, is our last NRP version supporting Python 2.7 and it is based on Ubuntu 18.04.

These two options are now available for both users accessing the NRP on-line and users wishing to install our sotware on their machines. Simply go to our Access the NRP page and don't forget to check the list of features for each version on our Release history and notes webpage!

SfN Global Connectome: A Virtual Event is the SfN brand-new, cross-cutting digital neuroscience event open to all neuroscientists and designed to facilitate scientific exchange across the globe and across the field, providing scientists at all career stages with opportunities to learn, collaborate, and connect.
The abstract of Dr Morin's presentation, "The HBP Neurorobotics Platform: from closed-loop neuroscience to embodied AI" is available here.

• Fluid simulation
• Plotting tools in the frontend
• Human-in-the-loop improvement in the Unity client
• Graphical experiment designer
• Web robot designer
• Improved SpiNNaker integration
• Support of HBP EBRAINS Sonata brain file format and SP6 Cerebellum model
• New models and experiment templates
• Python 3 ready

A screenshot of the video presentation.

Prof. Knoll introduced the Human Brain Project and the role that the project has in the European brain research. He then hihglighted the scientific goals that Neurorobotics is helping pursue within the project, especially thanks to the Neurorobotics Platform, which he defined as a new tool driving the virtualization of robotics and neuroscience. The video can be accessed through this link and it will be uploaded to the website of China Science and Tecnology Musem. Scientific American is a popular science American magazine and it is the oldest continuously published monthly magazine in the United States. The Chinese Scientific American was established in 1979.

One of RoBoBrain experimental setups in the NRP.

The title of the project is Whole-brain spiking neural network and rodent musculo- skeletal models integration for behavioral experimentation in the neurorobotics platform and it represents the very ambitious attempt to virtually embody a brain model of point neurons with realistic topographic connectivity, implementing closed loops and multiple regions interactions, to be then simulated at different scales. The proposed work builds upon current efforts that aim at providing the NRP with the ability to run large-scale NEST neural networks by distributing NEST simulations over multiple nodes of the Piz Daint cluster in EBRAINS.

Dr. Fabrice Morin presenting at the HBP Summit.

TUM senior researchers Dr. Morel and Dr. Morin introduced the NRP to the Summit audience with a presentation titled The Neurorobotics Platform: Embodiment as a bridge between model and function. After a brief introduction on the main concepts of embodiment, closed-lopp neuroscience and its connection to physical robotics, the presenters introduced the challenge of simulation and in silico experimentation, concluding with a list of the many and promising applications the NRP can lead to.

Dr. Porr discussing his feedback with the workshop participants.

The aim of this workshop was to strengthen and extend our understanding of the needs of current and prospective users of the NRP, so that the software development process of the NRP could be better aligned with those needs. The workshop was organised and coordinated by Dr. Yannick Morel and Victoria Neumann, and gathered 15 participants, both HBP members and external users. They exposed their expectations (and sometimes frustrations) in lively discussions over the two days of the meeting, providing invaluable insight into the value proposition of the NRP. In particular, we would like to thank Dr. Bernd Porr, from the University of Glasgow, for joining us and contributing with both feedback and ideas. A big thanks also to our dev team who actively contributed to an engaging and productive conversation throughout the workshop!