Janosch, would it be interesting for you to look beyond VSLAM applications? I can say that for social human-robot-interaction, the camera setup is one of the more complex pieces, and requires a fair bit of processing even for basic stuff. One needs to be able to move the camera around fast and slow, it must not be too big (which is, today, usually achieved by placing the sensor apart from the processing), and integration with an inertial sensor is definitely a bonus (one can do the vestibulo-ocular-reflex based on good joint sensors, but it's not the ideal thing). Furthermore, most of the heads I know of have abysmal optics (fixed focus, etc.) To support such things, your systems would, primarily, need to be able to cope with changes to the spatial relation of the cameras at runtime. cheers, Ingo On Fri, Jan 13, 2012 at 2:02 PM, Janosch Nikolic wrote: > Dear ROS users and roboticists, > > We (Swiss Federal Institute of Technology, ETH) are about to develop an open > Visual-Inertial low-cost camera system for robotics. The goal is a standard > piece of hardware that combines a monocular/stereo/multi camera and inertial > sensors (an IMU) providing synchronized, calibrated data in ROS. > > Such a piece of hardware could allow the community to focus more on the > development of novel algorithms and less on hardware and integration issues. > Our hope is that this will motivate young PhD students and researchers > around the world to make their VSLAM frameworks ROS compatible, ultimately > leading to 'plug and play VSLAM for everyone'. > > The system will be based on the latest generation of XILINX FPGAs > (Artix/Kintex-7 / Zynq, that's fixed) as well as a user programmable CPU and > provide at least a GigE port. We target real-time, low-power processing, and > the user should have access to raw data as well as pre-processed information > such as features and flow fields etc. > > We would like to invite you to give us feedback an thereby influence the > design of our system. We are extremely happy about any feedback you can > provide, more general feedback is as welcome as technicalities such as: > > - Cameras (number, configuration, resolution, HDR, global shutter, lens > type, etc.) > - General specs such as weight, size, power consumption, etc. > - Inertial sensor grade > - Interface to the host, API, etc. > - On-camera features (feature detection, dense optical flow, dense stereo, > etc.) > > Many thanks and best regards, > ETH - CVG > ETH - ASL > > _______________________________________________ > ros-users mailing list > ros-users@code.ros.org > https://code.ros.org/mailman/listinfo/ros-users -- Ingo Lütkebohle Bielefeld University http://www.techfak.uni-bielefeld.de/~iluetkeb/ PGP Fingerprint 3187 4DEC 47E6 1B1E 6F4F  57D4 CD90 C164 34AD CE5B