The human hand is one of the most complex pieces of natural engineering and the Shadow Robot Company has taken a truly anthropomorphic approach to building robot hands capable of manipulation.
The integration of the Shadow hand on the HBP Neurobotics Platform will provide a data-rich environment and food for thought around many possibilities and opportunities coming from neuroscience.
Shadow Robot Company has taken a truly anthropomorphic approach to building robots capable of manipulation. The Shadow Hand approximates the kinematics of the human hand – _from its overall size to the 20 actuated degrees of freedom along with a further 4 under-actuated movements for a total of 24 joints. The Shadow hand is fully sensed, with joint position, motor current and tendon force sensing across the whole hand, providing a data-rich environment for control technologies to use; its availability in simulations on the HBP Neurorobotics Platform will make it possible for existing and future Shadow Robot customers to easily explore learning paradigms and control models coming from neuroscience, something no other simulation platform can offer. The goal of this project is to explore the possibilities offered by neuroscience for motor control, especially with spiking neural networks and reinforcement learning.
The objectives and benefits of this project are:
- To evaluate the novel learning techniques and robot controllers from HBP researchers
- To enhance Shadow Robot´s current control system through embodied AI and brain-derived controllers using the Neurorobotics Platform
- To explore the use of spiking neural networks as a promising avenue for highly energy-efficient real-time control with neuromorphic hardware such as SpiNNaker
- To explore the design space of robot controllers
Collaboration with HBP
During this project, the Shadow – HBP collaboration would involve:
Developing a library of models in the NRP with Shadow’s advanced commercial-grade anthropomorphic hand models. Also, the platform’s value proposition will be enhanced with a more robust numerical simulation of grasping, as the need for reliable tools for numerical simulation of grasping is critical. Finally, the collaboration will result in increased visibility of the platform within the robotics community. In general, the work that it is going to be carried out will contribute to bridge the neuroscience and robotic communities.
NEWS from Michael Zechmair (M.Sc. TUM Department of Informatics), who is integrating the work of SHADOW into the HBP Neurorobotics Platform:
The Shadow Hand can now be simulated within HBP's Neurorobotics Platform.
The platform connects a physics simulator (either ODE or Simbody) to a variety of brains or neural networks, implemented e.g. via Tensorflow, NEST, Nengo. This allows for quick experimentation, testing and reconfiguration of neural networks, removing the need for time-consuming physical setups. Each Shadow Hand joint can be individually linked to both the input and/or output of the brain, giving the user a huge level of control over the link between hand and neural network. Contact based feedback can also be forwarded to the brain. Additionally, the NRP utilizes ROS, meaning any existing setup using this software can be ported to the platform.
Time frame: 2019 to 2020
Origin: HBP Voucher Programme