Plugin: Parallax

From Autopilot Wiki
Jump to: navigation, search
Parallax
Short Description Classes to control an array of stepper motors in a task to study depth perception from motion parallax
Compatible With Autopilot Version >=0.4.0
Git Repository URL https://github.com/auto-pi-lot/autopilot-plugin-parallax
Contributors sneakers-the-rat
Created By Jonny Saunders
Version 1
Used With Hardware Raspberry Pi Camera, LSM9DS1
Controls Hardware Stepper Motor
Plugin Type Hardware, Task, Children

hardware and task classes for studying Depth Perception from Motion Parallax. This is (i think!) unpublished work with Phil Parker in the Cris Niell lab, so more details will be forthcoming! They recently published this task in https://doi.org/10.1101/2021.09.29.462468v1

Platform

The platform is an array of stepper motors connected to vertical pillars on linear actuating rods that extends the Hardware class to control them. More information can be found in its docstring, but in short: the platform works by latching a certain set of pillars as active, then the step and direction signals are shared by all of them. To simulate independent movement, the active pillar latches are strobed (smartly, after a 2-d height command is given, pillars are raised and lowered until they reach the closest value, then deactivated, and so on.). The platform has both a position and velocity mode, and the velocity mode is used in closed loop with an LSM9DS1 IMU, along with the IMU_Orientation transform.

Parallax

The task itself, which I don't think I can say too much about! Except that it uses a combination of the IMU and live tracking with DLC (DeepLabCut) live to make a very fast but accurate estimate of the position of the subject wearing the IMU!

... more about "Parallax"
hardware.Hardware +  and tasks.Task +
sneakers-the-rat +
Classes to control an array of stepper motors in a task to study depth perception from motion parallax +
Phil Parker +  and Cris Niell +
Hardware +, Task +  and Children +
transform.geometry.IMU_Orientation +  and transform.image.DLC +