act12/results/Arm

Project Members:

Samir Menon Bryan Tripp

Project Overview:

Our goal was to control a robotic arm using a spiking neural circuit model in  Nengo. This was basically a prototype for a similar project on Neurogrid.

We integrated Nengo with a physics simulation of a simple 3-degree-of-freedom arm running in SCL (see  VIDEO). The Nengo neural model calculated the torques that were needed to smoothly move the robot to specified configurations, accounting for gravity and coupling between degrees of freedom.

We also started work on a more complex controller for a 6-DOF DENSO robot. We made a CAD model for the arm (see below) and found the control equations. We have yet to model these control equations in neurons, as they are much larger than the 3-DOF equations. However, since the workshop we have written a tool that generates a reasonable neural model automatically from control equations. This approach should allow us to build quite complex neural controllers in the future.

Attachments