Robotic Flying Squirrel

Research

• Biomimicry

• Market research

• Technology research

• Stake holder interviews

• Ideating

Design

• Snap fits

• Linkage design

• Design for Assembly

• Design for Additive Manufacture

• Iterating

The Brief

“Design a robotic gliding squirrel that can use solar power to climb trees to gather samples of fungi and record videos in the upper canopy of a rainforest. It can be designed for gliding from tree to tree to save energy. It is encouraged to take a robot design approach using recycled parts of bicycles, automobiles, or household waste as much as possible. This will allow local communities to easily repair or assemble robots for community use.”

the project

If the robot was to climb trees and jump and glide between them, it was to be necessary that the three aspects be designed in tandem. There were (and still are) a few examples of robots that are able to do at least two of the three, but often it was a case of a climbing robot being adapted to glide, meaning the robot ended up being good at neither. Therefore, to ensure the project’s success, the robot had to be able to climb, jump, land, and monitor together without compromise, and to be able to do all of this whilst being cheap to build from local resources to ensure its accessibility in remote parts of the world.

Research

It was found that existing robots were capable of jumping, or gliding, or climbing, or being built sustainability or some combination therein, but none were capable of doing all well. Therefore there was plenty to learn from existing projects, such as effective climbing or landing mechanisms, but scope to combine the best of existing designs to create something all the more capable.

Gliding and Landing

Flying squirrels are both efficient climbers and good gliders, so I decided to use biomimicry to try and influence my design. Through analysing their flight, it can be seen that they use their tail to adjust their orientation as they’re about to land, in order to both slow down and to land vertically (a process called “vortex shedding”).

I created an experiment to recreate behaviour robotically. An ultrasonic sensor was used to measure the distance to the landing surface, and PID control was used with this input to control the glider’s landing. A hair dryer was used to simulate airflow so the PID controller could be tuned, to ensure the glider would land vertically every time.

In this sequence of images, we can see the glider approach the landing surface, increase the angle of its rear wing in response, and successfully land on the surface vertically.

I used computer vision in Matlab to analyse the glider’s path and confirm that the glider was slowing down as it approached the landing surface. Through analysing the size of the bounding box output from the computer vision algorithm, four stages in the flight path emerged:

1. The bounding box gets smaller as the glider passes the camera into the distance

2. The bounding box size increases as the orientation of the glider changes in response to the tail angle increasing

3. This slows the glider down on, but its path continues ahead and getting smaller still into the distance

4. The glider hits the landing surface

Climbing

I explored many possible climbing mechanisms, attempting to model the linkages using Lego.

It was necessary that the climbing mechanism be as simple as possible, in order to save power and weight, so it was decided that the climbing be completed by a single motor, to control two linkages, one for each leg, contra-synchronised from each other. The linkage had to control the “feet” of the robot such that the feet would be able to hook themselves into the climbing surface to hold the robot against the tree, before unhooking themselves whist causing minimal damage to the tree and reaching further up to complete the climbing motion.

An adapted “Klann” linkage was chosen, as this gave the desired movement and orientation path for the feet. The linkage was first modelled in principle using Lego, where the length and position of each link could be quickly adjusted to optimise the movement, before being 3D printed for real world testing.

Ultimately it was found that the Klann linkage wasn’t quite ideal. It enabled the robot to move forwards slowly, but at least once per cycle the torque became extremely large, causing the motor to often stall and a huge power drain. Ultimately, more work is required to optimised the linkage for climbing.

Final Robot

The design was 3D printed and snap fit together for rudimentary real-world testing as a complete unit. The tolerances designed were not perfect however, so some had to be glued or taped together, but all the electronics worked as expected. There was not quite enough space for the battery inside the body due to all the wiring. All components snap fit together to make repairs and assembly quicker. The main body and wings of the robot are wrapped in a thin sheet or film. Standard, readily available off-the-shelf electronic components were used as much as possible, to make the robot as simple as possible. As few components as possible were used to help keep costs and weight down and to make the soldering process easier.