Controlling sequential movements with neural networks in neuromorphic hardware

In this project, we will implement a neural network on a spiking neuromorphic chip ROLLS [1] that will be able to plan and initiate serially ordered movements in a robotic agent. The student can bring his/her ideas what kind of sequential behaviour should be carried out. One example could be to learn and plan a path that includes specific places. Sequences can be learned with a serial order architecture that we implemented on the ROLLS in previous work. The ROLLS chip will receive sensory information from the Pushbot’s (https://inilabs.com/products/pushbot/) neuromorphic “artificial retina” camera DVS, which will be processed by the implemented neural network. This on-chip neural network will then control the Pushbot’s movement in a closed sensory-motor loop.

[1] Qiao, Ning, et al. "A reconfigurable on-line learning spiking neuromorphic processor comprising 256 neurons and 128K synapses." Frontiers in neuroscience 9 (2015).