1 of 14

Automating Forest Biodiversity Surveys

​

Team Green T and the Sushi Platter

(Andrew Gritsevskiy and Hanna Tuomi)

2 of 14

Motivation

  • “A new species is being discovered in the Amazon every two days” [1]
  • But studying the Amazon is hard!
  • Current methods are inefficient
    • Flying is really difficult and cannot get under the canopy
    • Ground travel is impeded by density
    • On-the ground sensors require set-up and man-power
  • What other ways are there to un-invasively monitor the Amazon?

​

3 of 14

How can we improve the monitoring of biodiversity in dense forest?

Create a robot to use the density to its advantage!

4 of 14

How do we do that?

  • Harness the tree density
  • Traveling in the understory layer eliminates non-tree obstacles but provides a clear view of the forest floor
  • Idea: Create an autonomous robot that moves/swings using the trees and captures measurements via sensors and footage

5 of 14

Minimum Viable Product

  • Focus the prototype on the motion that moves the robot from trunk-to-trunk using extendable arms and grippers
    • Start on an initial branch
    • Using computer vision, identify the next branch/trunk
    • The robot rotates and extends its other arm to grab onto the new tree
    • Once it grasps a new trunk, it retracts its other branch
    • These four steps are repeated for arboreal locomotion

​

​

6 of 14

Prototype Sketch

7 of 14

Build Phase

  • Simplified the robot
  • Wheels to support weight to allow for rotation
  • One side to demo the wrist motion
  • Other side to demo claw gripping

8 of 14

  • 12V relay circuit to control the linear actuators
  • Initially two sided (wrist +grip)
    • Issues with mobility + weight

9 of 14

  • Simplifying the design - prototype with one arm side and wrist motion

10 of 14

Computer Vision Testing

  • Plan: use semantic segmentation and parallax
  • Working model: HSV filter + edge detection

11 of 14

Computer Vision Testing

  • Plan: use semantic segmentation and parallax
  • Working model: HSV filter + edge detection

12 of 14

Final Design

13 of 14

Final Design

14 of 14

Thank you!

What are your questions?