Over the last couple of weeks I have read many news articles covering The Synergy Moon and its plans to use NVIDIA's Tegra 3 system for its rovers, The Tesla Prospector and The Tesla Survivor. The articles were also referring to my previous collaboration (ESA version) with the European Space Agency and my PhD research for the for the ITALK Project supported by FP7 Programme of European Union and coordinated by my supervisor Angelo Cangelosi.

I can confirm that the rumours regarding our use of NVIDIA's Tegra 3 system were correct, however, I noticed that publishers often confused CUDA with Tegra 3. We are proud to work with Tegra 3 next-generation systems featuring quadcore processors, 12 GPU processors and one additional companion core. We are of course aware of the harsh conditions in space and on the lunar surface and I can assure you that we have got specific plans how to tackle this and protect Tegra 3 system in our rovers. Tegra 3 combines all the benefits we would hoped for such as low power consumption, small size and low weight and fantastic processing speeds. Tegra 3 also includes 12 GPU (graphics processing unit), which are great for processing parallel problems. My whole PhD research has been very positively influenced by NVIDIA's technology and also by NVIDIA's staff who encouraged me to carry on with my efforts.

Couple of years ago I read an article that used NVIDIA's CUDA (Compute Unified Device Architecture) to implement neural networks running on GPU processors as opposed to standard CPU processors. The article inspired me a lot because it helped me to realise that GPUs are ideal for cognitive robotics research where we develop complex bio-inspired systems that are by nature parallel. I have started development of CUDA-based software named after constellation Aquila. Aquila has been very successful and helped me to demonstrate the power of NVIDIA's GPUs when used for cognitive robotics research. This led to couple of articles featured on NVIDIA's website and later to my visit to NVIDIA's headquarters where I did a presentation about my PhD research and involvement with The Google Lunar X PRIZE Synergy Moon team. Later, we received the Tegra 3 system and we are going to use it to develop state of the art AI system for the rovers.

NVIDIA has been a great inspiration and I have even started my own business (ProIS Group) that utilize the power of GPUs in the industrial sector. If everything goes well, The Synergy Moon team will launch two rovers controlled by Tegra 3 systems, which will be the first time ever when NVIDIA's products reached other celestial body. 

Press Coverage: 

Today, this site will hit a quarter of million visits. Thanks for popping over and if you have any feedback/suggestions please visit my guestbook and leave your comments. 

Olav Zipser, the pioneer of the freeflying, has recently returned from Moscow where he met with world-leading manufacturers of space suits. I have previously written article about Olav's aspirations to exceed the supersonic speed and his last visit to Russia was an important step towards reaching this goal because jumping from 40km altitude needs not only the mastery of the freeflying but also a space suit that will be custom made for Olav's body. The video below is a great summary of his new FreeFly Astronaut Project.

I would like to announce that we have just released a new version of Aquila. This is the first alpha version (v1.8a), which is much more stable that the previous versions. Apart from several bug fixes, this version has two major additions. The first improvement is in terms of stability where we replaced the previous rendering implementation causing occasional segmentation faults with a new one that is reliable and equally fast. The second addition is a new ERA (Epigenetic Robotics Architecture) module, which adds the functionality of the ERA architecture in a fully configurable manner rather than hard wired as in the MODI experiment tab. More advanced visual filters and free speech via yarp ports are also included.

There are few new developers working on the extension of the Kinect module to improve the precision of the control, make the interaction more fluid and add iCub legs control. Other developers are looking into developing an EEG interface to control objects in the simulator.

Aquila can be downloaded from iCub and ITALK repositories as well as from the main SourceForge repository that we use for the development. New developers are welcome to join the project via SourceForge and have their own branch in the repository. Aquila 1.7b manual is in the root directory and provides basic instructions on how to use different modules and also guidelines on how to compile the software. The 1.8a version of the manual is not yet finished but if you have any problems installing the latest version let us know.

The latest updates are always on Aquila's FaceBook page, click here to explore it.

Our humanoid robot iCub is going to be upgraded with new touch sensors and artificial skin. First, the robot's hands will be sent to IIT where skilled technicians will install new fingertips and the skin on each forearm. Later, we will send the whole robot for a complete upgrade. 

I have met today with our technician Bill, who also attended iCub technical training at IIT back in 2008. We have carefully disconnected all the wires and removed iCub's arms, which will be sent to Italy this Friday. In the meanwhile, we can keep using the remaining parts of the robot and carry on doing our research.

The video below shows how will our iCub look once the upgrade is completed. The video is also interesting because it shows the force control system utilising the new skin, which is a big advantage because one can apply force anywhere on the skin and the arm will move correctly. In contrast, the force control system that does not use skin assumes that one only applies forces at the end effectors, which are the hands/wrists.

 

In 2009, PhD researcher and my good friend Frédéric Delaunay from The University of Plymouth pioneered a novel design for a humanoid face. Fred is doing his PhD as part of The CONCEPT Project led by Tony Belpaeme. His design is cheaper and more robust alternative to other robot face technologies because it is based on retro-projected live-generated face animations. As you will see in the video below many current designs face serious problems when it comes to their price, maintenance or realness of face expressions and HRI (Human Robot Interaction)  experience. Fred's system uses a small laser projector that projects real-time face expressions on a semi-transparent face mask. These real-time animations are generated through his software based on an open source 3D content creation suite Blender.

Fred is currently extending the system so it could elicit rich and dynamical expressions that are not only based on the animated face expressions but also on their dynamics as well as the dynamics of head movements. I think we have a lot to look forward especially because many ongoing projects are using similar retro-projected faces. If you would like to stay updated on Fred's work then follow the LightHead page on Google+.