A July 25, 2016 article by Popular Mechanics,"The Robot Maid of the Future Might Teach Itself How to Do Chores," reports that TerraSwarm PI Pieter Abbeel is working with the non-profit artificial intelligence research company OpenAI on teaching robots to learn how to complete common household tasks. OpenAI hopes to use deep learning to enable robots to train themselves to complete manual tasks. This process has the potential to increase robots' ability to adapt to constantly changing real world environments, and substantially decrease the amount of robotic programming needed in the future.
PI Pieter Abeel brings his experience with robot learning and deep reinforcement learning to the project. Last year his research group successfully used algorithms to enable a robot to complete several simple tasks using a process of trial and error. Cited as a major milestone in AI, the process developed by PI Pieter Abeel's team more closely approximates how humans learn.
The TerraSwarm Research Center Blog covers news items about the TerraSwarm Research Center at http://www.terraswarm.org. The TerraSwarm Research Center, launched on January 15, 2013, is addressing the huge potential (and associated risks) of pervasive integration of smart, networked sensors and actuators into our connected world. The center is funded by the STARnet phase of the Focus Center Research Program (FCRP) administered by the Semiconductor Research Corporation (SRC).
Friday, July 29, 2016
Monday, July 25, 2016
Resarch Team Develops Tiny Computer with Enormous Potential
A July 15, 2016 Engineering.com article"Tiny Computer Has Enormous Potential," covers the work of University of Michigan Researchers including TerraSwarm PI David Blaauw and TerraSwarm PI Prabal Dutta. The article reports on the team's work developing the Michigan Micro Mote, an energy efficient computer built on a cubic millimeter scale. Outfitted with sensors for temperature, pressure, and imaging, this computer is able to network with other Michigan Micro Motes with a range of 2 meters.
The computers are built in layers using a Phoenix processor. With a standby power consumption of about 2nA, the Michigan Micro Mote is able to power its battery using artificial, ambient light and a 1mm2 photovoltaic cell. The solar module also serves as the mechanism to program the machine through a pattern of flashing light. Possible applications for these micro computers include medical devices, environmental monitoring, and surveillance. Going forward the research team hopes to increase the communication range for the Michigan Micro Mote and continue work on developing even smaller machines.
Wednesday, July 13, 2016
TerraSwarm PI Jan Rabaey Serves as Electrical Engineering Division Chair

Subscribe to:
Posts (Atom)