Robotics

From Webots to Realbots

ePuck robot

During this academic year, my Robotics module Comp2403 was taught through the Webots simulator. We used mainly the ePuck robot, a nice critter with a number of sensors including a color camera. Now we have in our posession one real, actual, physical, very expensive ePuck. This project will review various problems we solved using the Webot ePuck (line following, factory floor, maze-solving, object avoidance and basic navigation) and transfer these to the real ePuck. A cross-compiler is available, but we have no experience of using this.

Primary data would be drawn down from the simulated and physical ePuck. This could be used in a number of ways. One approach would be to critically compare the behaviour of real and simulated robot. Or you could choose to focus on the real ePuck and evaluate its performance in solving one or more problems.

Topics
Supervisor

Big Bad Robot

Big Bad Robot

Here you will design a controller for this Big Robot so it can navigate around Charles Hastings ground floor. You will need to research and select suitable sensors (laser, LIDAR, ultrasonic, collision, Kinect, color vision) and choose a system architecture, which may involve several microcontrollers (e.g., Arduinos) and even a laptop. You will design your software architecture with the hardware levels in mind.

Primary data would come from the robot. You would critically analyze its performance compared with the intended behaviour, for example did it manage to recognize and avoid stair-wells? Could it avoid humans walking around? Did it obey social distancing?

Topics
Supervisor

Educational Robot

Boebot

Here you will design, build and test a two-wheeled educational robot from raw components, to produce a cheap but sophisticated robot to solve problems such as maze following. You will select an appropriate micro-controller (e.g., an Arduino flavour), the drive system (stepper, servo or dc-motors with encoders) and a range of sensors. You will design the chassis and other mechanical components in CAD which will be laser-cut or 3D-printed for you. Then you will code a solution using a suitable IDE (matched to your microcontroller choice).

Primary data will be collected from the robot. Here you will compare its actual performance with the desired performance. A number of tests could be made; accuracy of moving on an arc of set radius, accuracy of navigation between obstacles.

Supervisor

Quadruped, Hexapod, Fish-like or Snake Robots

...

You've probably seen movie-clips of Spot the quadruped robot that can open doors. You may even have been on the Boston Dynamics Website https://www.bostondynamics.com/ with tons of interesting robots.

The controller for walking or swimming robots is often based on 'Central Pattern Generators' (CPGs), inspired by Biology, consisting of a small number of interconnected neurons. For a quadruped (think horse or dog) these CPGs can successfully reproduce various gaits such as trot, canter, gallop, pronk and bound.

This project will involve researching CPGs and constructing a simulator for a robot of your choice (biped, quadruped, hexapod, snake). The best way to code a CPG will be using Octave or Webots, then you may choose to import your code into a game-engine to produce a 3D-simulation. The engine of choice (mine) is UE4; here you will obtain experience in C++

Primary data will be collected from your simulator. This will be compared with data presented from research in numerous journal articles.

You may prefer to create a physical robot to run your code. This will be possible using servo-motors and 3D-printed parts which you will design in a CAD package.

Topics
Supervisor