October 21st, 2014:
At the moment the whole CoCoRo consortium is preparing for the final review. The decision which demonstrator will be presented was made, the AUVs are checked. In the last weeks a workshop took place at SSSA, Pontedera, Italy, to test and improve the algorithms that will be presented to the reviewers. At the moment a workshop for the final preparations takes place in the harbour in Livorno, Italy, to prepare the demonstration of the CoCoRo hardware in the open environment. The motivation of all partners is excellent, everyone is eager to present the results of this astonishing research project.
June 12th, 2014:
The 2014 IEEE International Conference on Robotics and Automation (ICRA) has been held at the Hong Kong Convention and Exhibition Center, Hong Kong, China, from May 31 to June 5, 2014.
The conference theme was “Robotics and Automation: Technologies Enabling New Economic Growth” reflecting the growing spectrum and recent developments in robotics and automation around the world.
The SSSA group proudly presented the paper “Mechatronics design of a miniature underwater robot for swarm operations” that describes the design of the Jeff robots of the CoCoRo project. The presentation topics have been the mechanical design of the platform with the description of custom high efficiency solutions for propulsion and the docking capability for power and information data exchanging. The project received compliments for the great job and has been approached by potential partners for future underwater researches.
April 28th, 2014:
In the CoCoRo project we are developing a communication-less shoaling algorithm. This algorithm works with just the (analogue blue-light) distance sensors on the AUVs and allows the AUVs to move together as a group. The algorithm is used in a purely reactive controller (i.e., without memory): Each AUV estimates the distance and direction of other near AUVs and reacts on these AUVs. These position estimates can fall into different zones around an AUV. The parameters for the sizes of these zones and for the reactions of the AUV are evolved in the simulator because they are highly interactive and hard to guess right. An example of an evolutionary run (10 swarms are evaluated per generation, for 2000 generations) can be seen in the picture to the right.
The figure shows that for the first ~500 generations only solutions with a fitness of up to 100k points were found (AUVs stay together, but do not align). After the ~500th generation new solutions with up to 450k fitness points were found. After the ~1000th generation slight mutations of these solutions lead to a more stable, but worse solution (300k fitness points, middle band) and to the best found solutions (550k fitness points, upper band: AUVs stay together, align, and move around).
March 24th, 2014:
From March 10th to March 14th the CoCoRo project proudly presented its largest swarm of interacting autonomous underwater vehicles at the CeBIT 2014 in Hannover.
For our very small 12 squaremeters booth we organized a large aquarium (2m x 1m x 80cm), filled it with 1500 litres of water and some gravel and put our Lily swarm in there.
As expected, our blinking underwater robot swarm attracted quite a lot of visitors who were eager to ask us about the function and the purpose of our swarm.
Our project was also featured prominently in the CeBIT press releases and also attracted multiple international camera crews for interviews.
One of our main goals at the CeBIT was also quite successful: We were approached by more than fifteen potential industrial partners (ranging from offshore oil industry to aerial marine rescue robots) for future follow-up projects.
March 24th, 2014:
This second blog entry about out CeBIT 2014 shows a better view of our aquarium and the Lily swarm inside.
In this picture you can see that four Lilys have successfully aggregated above a "point of interest" at the floor of the aquarium and have thus sloved a simulated real-world problem:
Our swarm of Lilys starts distributed over the whole aquarium in search of a sunken object (e.g., a black box of a crashed airplane) whose exact location is unknown.
In order to find the black box, our Lilys randomly move through the aquarium until one of them, by chance, is directly above it and thus can locate the sunken object.
This Lily then starts to emit blue-light and thus signals other near Lilys to come towards it. These near Lilys also emit the blue-light signal which leads to a fast aggregation above the object.
In a real-world scenario, the (larger and more robust) marine AUVs could then work together and, for example, use their grippers to collectively lift the black box to the water surface.
January 27th, 2014:
In the second week of January 2014 the CoCoRo consortium had a workshop at SSSA in Pontedera, Italy. SSSA prepared an indoor pool for experiments to test the first autonomous Jeff AUVs. First tests included the adaptation of the Lily depth controller to the Jeff depth controller. Both AUVs can change their buoyancy by changing their volume: Lily uses a cylinder the moves in- and outward whereas Jeff uses an elastic membrane that is pushed in- or outward. The calibration of such a depth controller is quite complicated, but by the end of the workshop we managed to give the Jeff AUV the capability to autonomously sink, rise, or keep its depth.
The picture shows (part of) a swarm of 5 Lily AUVs with 4 Jeff AUVs diving in the pool. A video can be found in the media section of the CoCoRo homepage.
December 16th, 2013:
Starting from August, the SSSA team is mass producing the Jeff AUVs. After receiving all the electronic boards from the partners and manufacturing of all the required mechanical components, the assembly of 20 Jeff AUVs is now on-going. The assembly involves over 200 components for each robot due to its multiple features. Indeed Jeff is capable of 3D swimming, docking to a surface station for battery recharging and it is equipped with multiple sensors for communication (blue light, pressure wave, potential field), obstacle avoidance and shoaling.
The picture shows the mass production of Jeff with all the main parts of the shell ready for the final assembly. The final version of the robot is shown as well.
November 27th, 2013:
Throughout October and the early weeks of November, the University of York branch of the project have been polishing off the functionality of the blue light communication boards, finally flashing the boards with the finalised firmware. These boards (all 108 of them!!) are now in transit (maybe have arrived) to Pisa, where they will integrated the Jeff platform, forming one of its major communication channels. These new blue light boards will bring omnidirectional robust communication within a swarm of Jeff platforms, allowing real world testing of the hypotheses brought forward by this project.
Shown is a picture of the flashing (and subsequent LED flashing) of one of the prototype blue light boards, witch is almost identical in function and from to the ones that the Jeff will be using.
November 11th, 2013:
At the ECAL 2013 (12th European Conference on Artificial Life) conference in Taormina, Italy, Payam Zahadat (Karl-Franzens University Graz) presented her work with the title "Social Inhibition Manages Division of Labour in Artificial Swarm Systems", by Payam Zahadat, Karl Crailsheim, Thomas Schmickl. The talk was about an algorithm for division of labour in artificial swarm systems. The algorithm was inspired by mechanisms of behavioural development in honey bees and targeted for robotic swarms. She presented the results of implementing the algorithm in virtual and simulated robot swarms representing flexibility of the swarm to changes in the number of agents and task-demands. The Proceedings of the conference, including the paper of Payam Zahadat regarding division of labour in artificial swarms can be downloaded at the MIT press.
September 18th, 2013:
At the ECAL 2013 (12th European Conference on Artificial Life) conference in Taomina, Italy, Ronald Thenius (Karl-Franzens University Graz) presented his concept of Emotional Artificial Neural networks (EMANN). In his talk "EMANN - a model of emotions in an artificial neural network" Thenius described the necessity for the development of a bottom-up model of emotional processes in a brain, based on the processes interaction of neuromodulators (e.g., hormones) and neural cells. Based on the theoretical constraints defined by Jean-Marc Felous, Thenius developed a model of neural cells, hormone gland cells and hormones. In the presented model these hormones have the ability to interact with the neural cells and change abilities of the cell (e.g., the ability to produce spontaneous activity). The presentation lead to a series of very fruitful discussions (on- and off-stage) regarding possible applications of EMANN and possibilities of comparison with top-down models of emotion and its influence on the brain. The Proceedings of the conference, including the paper of Ronald Thenius regarding EMANNs can be downloaded at the MIT press.