A couple of years ago, I regaled you with the story of MABEL, a robot designed to study bipedal locomotion. Among other things, MABEL aims to achieve a better compromise in "speed, stability, agility, and energy efficiency" than other bipedal-robot designs. (I'm quoting Jessie Grizzle, a professor of electrical engineering at the University of Michigan and one of the masterminds of the MABEL project.)
To acquire data from sensors, compute control actions, and output commands to actuators, MABEL uses a realtime computing and data acquisition environment based on the QNX Neutrino RTOS. The software framework for the control system is based on RHexLib, which was originally developed for RHex, another QNX-based robot.
If you're in the mood for a deep dive on MABEL, check out the April 2011 edition of IEEE Control Systems magazine — the article runs an impressive 25 pages. If, however, you'd like something a little lighter, I suggest the three following videos:
MABEL walking on an uneven surface
MABEL walking like a human, with "heel-strike and toe-off"
MABEL, before and after correction of a hardware bug