Simulating intelligent agents within realistic environments.
- In silico neuroscience: Improve and refine novel models of brain functions and architecture
- Co-design of robotic body and controller: Design tools for advanced robotics
- Cognitive robotics: Robots with enhanced decision-making capabilities and contextual awareness, for use in autonomous exploration (deep sea, space), search-and-rescue in disaster areas, and more
- Testing and verification of robotic behaviour: Implement testing, even under safety-critical situations
What is the EBRAINS Neurorobotics Platform?
The Neurorobotics Platform, NRP, is an integrative simulation framework that enables in silico experimentation and embodiment of brain models inside virtual agents interacting with realistic simulated environments. The NRP is an open-access, open-source tool, that users can use with different scientific backgrounds and levels of expertise.
Thanks to its direct connection to both Neuroscience research in the Human Brain Project and the EBRAINS HPC infrastructure, the NRP is the best available platform for scientific exploration and development of neuromorphic computing applications.
EBRAINS Neurorobotics Platform: An integrative simulation framework
Drawing upon the potential of both neuroscience and Artificial Intelligence in robotics, the NRP allows users to observe, analyse and test the emergence of behavioural patterns in virtual agents controlled by state-of-the-art models of brain architecture and functions. The result is an unprecedented approach to simulation, in which theoretical brain models can be checked against data-driven models thanks to simulations that take into account the dynamics of both the environment and the agent itself. This approach is essential to develop novel robotic control technologies that are robust, adaptable, and comparable to biological systems.
The NRP is the only open-source platform with concrete ties to neuroscience and the only one aiming to run large-scale simulations of spiking neural networks.
Embodied in silico neuroscience
Simulation is used in neuroscience as a tool to improve and refine brain models. By enabling the creation of closed action-perception-cognition loops in embodied settings, the NRP allows neuroscientists to observe and analyse emergent behavioural patterns in virtual agents controlled by specific models of brain architecture and functions. This gives them the extraordinary opportunity to test and refine these models iteratively without the many approximations entailed by simulating the brain decoupled from the body.
Neuroscientists are using the NRP for:
- Replicating a sensory-motor experiment in simulation
- Iterative testing and refining of functional brain models
- Implementing learning paradigms in large-scale parallel simulations
The NRP for roboticists and AI specialists
With the NRP, testing procedures that can otherwise be costly or difficult to implement on a physical setup - like testing of safety-critical situations or scenarios that could result in damage to the robot - can be performed without incurring any risk.
Inspired by biological architectures and with the possibility to reproduce any desired environment or experimental condition in the virtual space of the NRP, roboticists using our platform can test and develop new robotic controllers that will bring us the next generation of safe and adaptable robotic systems with enhanced cognitive abilities, such as contextual awareness and decision-making.
The NRP Docker technology is ideal for distributed / parallel learning and optimisation. Roboticists are using the NRP for:
- Generating synthetic data for AI research
- In silico evaluation of the functional performance of control architectures
- Implementing learning/optimisation frameworks in embodied simulation