intelligence Archives – THINK Magazine

Enter the swarm

Author: Jean Luc Farrugia

Once upon a time, the term ‘robot’ conjured up images of futuristic machines from the realm of science fiction. However, we can find the roots of automation much closer to home.

Nature is the great teacher. In the early days, when Artificial Intelligence was driven by symbolic AI (whereby entities in an environment are represented by symbols which are processed by mathematical and logical rules to make decisions on what actions to take), Australian entrepreneur and roboticist Rodney Brooks looked to animals for inspiration. There, he observed highly intelligent behaviours; take lionesses’ ability to coordinate and hunt down prey, or elephants’ skill in navigating vast lands using their senses. These creatures needed no maps, no mathematical models, and yet left even the best robots in the dust.

This gave rise to a slew of biologically-inspired approaches. Successful applications include domestic robot vacuums and space exploration rovers.

Swarm Robotics is an approach that extends this concept by taking a cue from collaborative behaviours used by animals like ants or bees, all while harnessing the emerging IoT (Internet of Things) trend that allows technology to communicate.

Supervised by Prof. Ing. Simon G. Fabri, I designed a system that enabled a group of robots to intelligently arrange themselves into different patterns while in motion, just like a herd of elephants, a flock of birds, or even a group of dancers!

I built and tested my system using real robots, which had to transport a box to target destinations chosen by the user. Unlike previous work, the algorithms I developed are not restricted by formation shape. My robots can change shape on the fly, allowing them to adapt to the task at hand. The system is quite simple and easy to use.

The group consisted of three robots designed using inexpensive off-the-shelf components. Simulations confirmed that it could be used for larger groups. The robots could push, grasp, and cage objects to move them from point A to B. To cage an object the robots move around it to bind it, then move together to push it around. Caging proved to be the strongest method, delivering the object even when a robot became immobilised, though grasping delivered more accurate results.

Collective transportation can have a great impact on the world’s economy. From the construction and manufacturing industries, to container terminal operations, robots can replace humans to protect them from the dangerous scenarios many workers face on a daily basis.

This research project was carried out as part of the M.Sc. in Engineering (Electrical) programme at the Faculty of Engineering. A paper entitled “Swarm Robotics for Object Transportation” was published at the UKACCControl 2018 conference, available on IEEE Xplore digital library.

https://www.facebook.com/ThinkUM/videos/493872941442263/

#research ai artifical automation intelligence internet internet of things iot issue 28 jean farrugia Jean Luc Farrugia malta robot robotics swarm swarm robotics technology

Escape the (Virtual) Room!

Virtual Reality (VR) has created a whole new realm of experiences. By placing people into varied situations and environments, VR enables them not only to explore, but to challenge themselves and gain skills in ways never thought possible. With applications in medical and psychological treatment, VR is now being used to train surgeons, treat PTSD, and to help people understand what it’s like to be on the autism spectrum. The key to this application is VR’s ability to immerse its users.

Many agree that immersion needs two key ingredients: a sense of presence and interaction with the environment. Interaction comes in three main forms. Selection is about differentiating between items in the environment. Navigation allows travelling from one point to another and observing the environment. Finally, manipulation lets users grab, move and rotate selectable items. In addition to this, VR applications need a setting. Supervised by Dr Vanessa Camilleri and Prof. Alexiei Dingli, I chose to use escape rooms (adventure games where multiple puzzles are solved to leave a room) to experiment with these interaction techniques.

I used escape rooms because they’re highly interactive and naturally immersive systems. And since interaction isn’t a one-size-fits-all scenario, I also applied procedural content generation (PCG) techniques to create the escape rooms themselves.

People selected items using a reticle, a small circle in the middle of the screen which expands or contracts to indicate which objects they could interact with. They navigated the space by looking around through the VR headset and moving their joystick. They manipulated puzzles from a separate screen which I layered on top of the escape room. This allowed them to inspect objects to their heart’s content, while also reducing the amount of clutter in the room.

Since there was no previous work in PCG escape rooms, I had to pave my own way. I used a genetic algorithm, a machine learning algorithm that mimics evolution in biology to select the best solution to a problem, to determine which puzzles and items would be placed in the escape room. I also programmed the game to create the rest of the room, placing floors, ceilings, and everything else that the algorithm didn’t consider. This made the space look like it had been made by an actual person, despite being created through AI.

From the results gathered, most people found that the system allowed them to explore the VR environment in a very natural way. Players said that the room’s generated interaction was consistent, reliable, and fun.

Understanding immersion is critical for VR’s future applications. If we can help people hone these techniques by creating a few games along the way—so be it!

This research was carried out as part of a Bachelor of Artificial Intelligence at the Faculty of ICT, University of Malta

Author: Natalia Mallia