Robohub.org
 

The Year of CoCoRo Video #30/52: Combined scenario number one

by
29 July 2015



share this:

CoCoRo30The EU-funded Collective Cognitive Robotics (CoCoRo) project has built a swarm of 41 autonomous underwater vehicles (AVs) that show collective cognition. Throughout 2015 – The Year of CoCoRo – we’ll be uploading a new weekly video detailing the latest stage in its development. This week there are two videos. The first shows a computer animation of  our “combined scenario #1.” The second shows how this scenario was performed by real robots in the water.

In CoCoRo, the search for a sunken metallic object on the waterbed was dealt with in several scenarios. The first video is a computer animation illustrating the phases of the most complex, “scenario #1:”

  1. The base station at the water surface moves into the habitat, with docked Jeff robots and a swarm of Lily robots confined to it.
  2. The Jeff robots are released. They sink to the ground and search the habitat.
  3. As soon as the first Jeff robot has found the target, it recruits other Jeff robots to the location via blue-light LED signals, leading to a positive feedback loop that attracts more and more Jeff robots to the area.
  4. Meanwhile, the Lily robots build a chain that connects the surface station to the aggregated Jeff robots on the ground.
  5. Through this “relay chain,” information (blink signals or other data) can be sent from the ground swarm to the surface station (and human operators) and vice versa.

Scenario #1 combines many of the individual algorithms and functionalities shown in previous videos in “The Year of CoCoRo.”

The second video shows how this scenario was performed by real robots in the water. It is noteworthy that all the phases in the scenario were executed in one single run and not split into different experiments. It was important for us to demonstrate that, using the modular algorithmic design of CoCoRo, it is possible to build up complex behavioral programs to solve composite tasks.

To date, in bio-inspired and swarm robotics, there have either been simple algorithms to enable robots to perform basic tasks or complex scenarios achieved by very complicated and specifically written software. To our knowledge, this is the first time in swarm robotics that several very simple signal exchange patterns and behavioral responses, triggered by simple signals within the environment, are combined with an elementary piece of code that allows the robot swarm to perform a very sophisticated behavioral program. There is no recourse to global knowledge, mapping, ego-positioning, image tracking or computer vision within this autonomous swarm. There are only: a few blue-light blinks, photoreceptors to receive these, random-walk programs, communication-free shoaling-behavior for the relay chain, radio-frequency pulsing and a simple internal-robot compass. This is how swarm robotics should be: simple, robust, flexible, scalable and – most importantly – it works!



tags: , , , , , , ,


Thomas Schmickl is an Associate Professor at Karl-Franzens University, Graz, Austria, and a lecturer at the University for Applied Sciences in St. Pölten, Austria.
Thomas Schmickl is an Associate Professor at Karl-Franzens University, Graz, Austria, and a lecturer at the University for Applied Sciences in St. Pölten, Austria.





Related posts :



Countering Luddite politicians with life (and cost) saving machines

Beyond aerial tricks, drones are now being deployed in novel ways to fill the labor gap of menial jobs that have not returned since the pandemic.
04 December 2022, by

Call for robot holiday videos 2022

That’s right! You better not run, you better not hide, you better watch out for brand new robot holiday videos on Robohub!
02 December 2022, by

The Utah Bionic Leg: A motorized prosthetic for lower-limb amputees

Lenzi’s Utah Bionic Leg uses motors, processors, and advanced artificial intelligence that all work together to give amputees more power to walk, stand-up, sit-down, and ascend and descend stairs and ramps.

Touch sensing: An important tool for mobile robot navigation

Proximal sensing often is a blind spot for most long range sensors such as cameras and lidars for which touch sensors could serve as a complementary modality.
29 November 2022, by

Study: Automation drives income inequality

New data suggest most of the growth in the wage gap since 1980 comes from automation displacing less-educated workers.
27 November 2022, by

Flocks of assembler robots show potential for making larger structures

Researchers make progress toward groups of robots that could build almost anything, including buildings, vehicles, and even bigger robots.
25 November 2022, by





©2021 - ROBOTS Association


 












©2021 - ROBOTS Association