Neurorobotics Approach to Learn On the Fly
In this project, we will explore a Neurorobotics approach to develop robotic controllers to scenarios which behavioural responses must be fast and precise whilst limited to strict energy constraints. This will be accomplished by the use of Evolving Spiking Neural Networks (ESNN) models.
Results from this project would be applicable to robotics in search and rescue missions, dangerous scenarios, critical operations and a plethora of human-robot-interaction scenarios, where enhanced autonomy, robustness and rapid responses are of vital importance.
In this context, ESNN have shown substantial advantages in comparison to traditional ANNs and other ML techniques. ESNN are more biologically realistic than its counterparts and take into account single neuronal membrane dynamics, connectivity, and plasticity. As a consequence, ESNN are faster, more robust, require less energy, and capable of online learning.
Traditional ML methods struggle to deal with huge streams of data, given that storage of whole datasets is unfeasible and, most notably, data distribution may change.
Moreover, several robotics applications require learning on the fly, thus algorithms have to be adaptive and sensitive to eventual changes in data.
Therefore, most critical robotics applications would benefit directly from using this novel approach of spiking neural networks, which are mandatory to solve complex tasks and support design requirements such as autonomy and robustness.
Lobo, J. L. et al. (2020) ‘Spiking Neural Networks and online learning: An overview and perspectives’, Neural Networks. Elsevier Ltd, 121, pp. 88–100. doi:10.1016/j.neunet.2019.09.004.
Tieck, J. C. V. et al. (2020) ‘A spiking network classifies human sEMG signals and triggers finger reflexes on arobotic hand’, Robotics and Autonomous Systems. Elsevier B.V., 131, p. 103566. doi:10.1016/j.robot.2020.103566.
Taherkhani, A. et al. (2020) ‘A review of learning in biologically plausible spiking neural networks’, NeuralNetworks. Elsevier Ltd, 122, pp. 253–272. doi:10.1016/j.neunet.2019.09.036.