Factory swarm

4 min read

The field of experimental robotics is arguably one of the most exciting in all of engineering research. Certainly it captures the public’s imagination like few other sectors, as testified by the number of viral online videos showing flying robots performing incredible stunts and even, in one instance, playing tennis.

The next stage in the evolution of robotics is likely to see robots working together in teams. The advantages are, literally, numerous - the ability to perform more complex assignments far more quickly and with inherent redundancy in case of failure of individual robots.

That’s exactly what Prof Anibal Ollero at the University of Seville is aiming to achieve in leading two related EU FP7 projects with a combined budget of more than €10m (approximately £8.4m). The larger project — Aerial Robotics Co-operative Assembly System (ARCAS) — envisions fleets of autonomous aerial vehicles co-operating to build new structures part by part, as well as inspecting existing infrastructure and, when required, dismantling it.

The craft needs to be reactive and come up with plans on the hoof
Approaching sites form air opens up entirely new possibilities

Approaching sites from air rather than land opens up entirely new possibilities, such as bridge building in challenging, mountainous terrain, industrial inspection of thousands of kilometres of oil and gas pipes, and even, in the future, taking apart retired satellites and space junk.

The project is aiming to demonstrate this capability using a range of robotic rotorcraft equipped with multi-joint manipulator arms for carrying and positioning various parts of structures. ‘We believe we will be the first in the world with this type of aerial manipulation capability, which includes perception and planning both outdoors and indoors, when the project comes to fruition in 2015,’ said Ollero. ‘There are systems that appear to do this, but they are designed for very specific, pre-defined indoor tasks, often using magnets to pick up and drop. What we want to develop is a fleet of robots with multifunctional arms with up to seven degrees of freedom that can adapt to many different applications and scenarios.’

The project aims to demonstrate co-operative assembly in an indoor test-bed using up to 10 autonomous quadcopters with manipulators. This will be followed by more advanced outdoor experiments with autonomous helicopters equipped with industrial arms, in addition to a ground-based test-bed to develop space applications.

Already, the team has first-stage design prototypes for these two craft, which are now in testing. Perhaps the most difficult task now though will be to imbue the craft with perception, planning and co-ordination capabilities.

Even when performing simple manoeuvres indoors, the microprocessers of quadcopters send signals to the motors hundreds of times each second to make tiny adjustments to pitch, yaw, and so on, to keep the craft airborne. Add to that other capabilities such as co-operative building and you are talking about quite an advanced flying brain.

The craft needs to be reactive and come up with plans on the hoof
The craft needs to be reactive and come up with plans on the hoof

‘You have to take into account all the local conditions, including wind if outdoors, as well as the complex aerodynamics that occur when a craft is very close to a surface,’ said Ollero. Prior to a mission, the swarm will be given a briefing complete with 3D maps and other telemetry. But the craft will

also need to be reactive and come up with plans on the hoof, based what they actually find at the site.

‘There has to be contingency; you have to be able to adapt; for example, if one of the parts falls down, you have to re-plan the assembly,’ said Ollero. The first applications of the core technology are likely to be limited to situations where it is currently dangerous and particularly difficult for human pilots to fly, which is the focus of Ollero’s other EU project: EC-SAFEMOBIL. That will include monitoring in disaster zones, with autonomous fleets capable of taking off from and landing on mobile platforms such as ship decks.

Perhaps though, in the not-too-distant future, it will become a common sight to see our growing city skylines dotted with airborne insectoid builders, busy day and night.

While rotorcraft are certainly one of the more popular research tools for swarm robotics, a team at Southampton University is considering a radically different approach.

The university, a leader in autonomous aerial vehicle research, recently began work on MAVIS (Massive Atmospheric Volume Instrumentation System) — a new project funded by the Engineering and Physical Sciences Research Council.

The aim is to send a weather balloon-borne ‘mothership’ high into the upper atmosphere, which would then release a swarm of autonomous gliders on a downward trajectory picking up the scent of pollutants and atmospheric markers as it goes. Such a system would be a huge boon for environmental scientists, providing high-quality data over a widespread area, often around hazardous locations such as active volcanoes.

One of the biggest challenges of sending out a fleet of up to 20 autonomous gliders at these altitudes is safety, explained project lead Dr András Sóbester. ‘We are trying to make a system that’s as innocuous as possible by reducing the weight and kinetic energy of the vehicles. Essentially we’re talking about no more than a few tens of grams of foam with some electronics in the middle.’

The craft needs to be reactive and come up with plans on the hoof
The robotic rotorcraft is equipped with multi-joint manipulator arms for carrying and positioning various parts of structures

Nevertheless, the ‘brains’ of the gliders will have to be relatively sophisticated to ensure that the swarm takes the most efficient path to ensure maximum data capture. ‘We could be looking for some kind of narrow pollutant plume in a very large block of atmosphere, and so one of the glider’s sensors might pick up traces of whatever we’re looking for, bringing the rest of the swarm in that direction as well,’ said Sóbester. ‘However, it’s a complicated balancing act: you don’t want to throw all your resources at the first thing you find, but equally you don’t want to spend too much time exploring uninteresting blocks of airspace,’ he added.

Part of that collective decision-making process will be based around evolutionary algorithms. Much the same as biological evolution progressively hones advantageous offshoots and stymies dead ends, these algorithms will select the best possible flight paths. Having already demonstrated launching of single craft from a weather balloon, the team hopes to be able to release multiple craft by ‘the spring of 2014’, according to Sóbester.

With the project partners — mainly the Scottish Association for Marine Science as well as the British Antarctic Survey, the Met Office and the National Centre for Atmospheric Science — Sóbester and team will then look for specific applications. ‘Of course, we can spend as much time as we like playing with aircraft as that’s what we like doing,’ said Sóbester. ‘But we are also eager to produce something that’s of genuine use to scientific missions.’