BRD18: Building applications with Braindrop—a novel neuromorphic chip for embodied perception and action
- The purpose of this workgroup is to give participants an opportunity to build applications with Braindrop. Braindrop is a new mixed analog/digital chip that implements the Neural Engineering Framework (NEF) and is programmable with the Nengo development environment.
- Sensory-motor coordination: The system must identify a visual target and track it (e.g., track a rolling ball with a camera and catch it). Demonstrates sensory recognition, motion processing, and motor co-ordination.
- Perceptual decision-making: Recent work in neuroscience has shown that decision-making is contextually controlled (Mante et al., 2013). This platform is ideal for building a model and testing it over hundreds of trials to reproduce the dynamics observed in monkey recordings.
- Adaptive control: A robust point-to-point controller must be able to adapt to variability in the system being controlled while online. Neural systems are ideal for computing the kinds of functions required of such controllers.
- Neuromorphic computing frameworks: Students will identify a small set of neural functions to compute (e.g., an integrator, a controlled oscillator, an interpolation function), and evaluate various frameworks’ strengths and weaknesses on those functions.
- Braindrop boards, quadcopters, Lego robotics
- The Neural Engineering Framework (NEF) – Chris Eliasmith
- The Nengo Simulator – Terrence Stewart
- Mapping Applications onto Braindrop’s Architecture – Alex Neckar
- Analog Computation cum Digital Communication – Kwabena Boahen