Robots navigate mazes in "DisCo" engineering course

Tuesday 24 May 2022

Watch as a fleet of robot rovers autonomously navigate mazes and their human creators root them on during "Final Rover Testing" week at the ANU School of Engineering:

The robots, designed and built by first-year students taking part in a course called Discovering Engineering, or DisCo for short, were programmed to pilot themselves using sonar sensors to detect and avoid obstructions.

"I was expecting to have to start with a bunch of Maths and project planning before actually getting to start working on any kind of project," said Matthew Britton. "So being handed a very open ended project right at the start was a lot of fun, because it allowed us to be very creative with what we did, while also learning how to manage a group and our time."

The students worked in teams and were supplied with a "bag of electronics", including ultrasonic sensors, wheels, and motors. An electronic prototyping platform called Arduino allowed them to upload coding from a laptop to a small piece of hardware that served as the robot's "brain".
 
"The Arduino would read the received signals from the sonar sensor," explained Yingjie Wang. "If the received signals implies that the distance between the obstacle in front of the sensor and the rover is shorter than a certain distance, the Arduino would communicate to the motors to either make the rover stop or to turn a corner."
 
The final two grading levels for the course involved navigating a known maze, and navigating an unknown maze. The rovers needed to complete the mazes in under 3 minutes, with points deducted for touching the walls.
 
Out of 59 robots, 19 made it all the way through the known maze, nine of these with perfect scores. Only nine robots were able to navigate the unknown maze, three with perfect scores. Forty-nine robots could solve at least part of the known maze, while 18 solved at least part of the unknown maze.
 
No two rovers were identical, and few looked alike despite the fact the teams had been supplied with the same instructions and component parts. 
 
"Big, important engineering problems tend to be ill-defined and open-ended," said course convenor Dr Fiona Beck. "This means they can’t be solved by applying known formulas, and they could have multiple acceptable solutions."
 
Britton said the biggest challenge for him was learning C++, the programming language used to shape the robot's brain. A servo motor turned the ultrasonic sensors (robot's eyes) left and right to measure the distance of obstacles to the right, 45 degrees to the right, straight ahead, 45 degrees to the left, and to the left.
 
"From there we applied some logic to these measurements," Britton said. "If the rover was too far from the left hand wall, you turn the rover left by adjusting which motor is powered, what direction it rotates and how long it is allowed to rotate."
 
If the rover detected that there was a wall on its left, it would go straight. If it detected nothing on its left, it would turn to the left. And if it detected a wall on its left and a wall in front of it, the rover would turn right, Britton explained.
 
"The core of our algorithm is that the rover does its best to stay within 7 cm of the left hand wall, because as long as the rover can do this, it can complete any maze," he said.
 
Wang praised her teammates, who chose to name their robot "Tom," as wonderful individuals who were "collaborative and goal focused". She also praised the tutors and course convenors, Beck and educational systems officer Jenny Simmons.
 
"They did an excellent job scheduling it so that we were not thrown into the deep end all at once. Rather, the skills that we need are taught to us gradually in response to the milestones," Wang said. "In addition, new technical skills are put into practise every week in tutorials where we are also able to seek help from peers and tutors."
 
Britton said that his team's robot made it all the way through the known maze, but only a quarter of the way through the unknown maze. He noticed that other designs that "used more than one ultrasonic sensor seemed to fair better."
 
Wang said her team has similar results.  "My team's rover was successful in the known maze," she said. "But unfortunately only passed the first one fourth of the unknown maze because one of the motors fell off."

 

Image gallery

Updated:  10 August 2021/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing