Cornell Ranger robot: This robot has walked 9 km under its own power, without falling. This semester we will refine and improve its hardware and controls. The robot requires some additional mechanical and electrical modifications, and extensive control-software development, using C++. We are also working on software integration (MATLAB, LabView) of measurement data and video records from robot tests, for rapid development and problem diagnosis. With the software in place, we will be walk-testing the robot and developing new, more power-efficient and robust walker controllers.

Mammals have developed sensorimotor controls in which a portion of the sensory processing and motor control takes place below the level of the brain, in the spinal cord or the extremities. For a control system with multitudes of sensors and actuators, this results in lower bandwidth and transmission speed requirements. Complex robots face a similar challenge, and we would like to develop a solution based on similar ideas. It should be modular and highly power-efficient, suitable for battery-powered robots with many sensors and controlled joints. Designs are in progress for a network of miniature low-power ARM processor boards, linked to each other and a central brain processor by a CAN network. Students would work on sensor, motor control, and microcontroller circuit boards, develop microcontroller software (C, C++), and conduct tests of the network and individual boards.

What are the available techniques for reliable control of complex robotic systems? What can be done to alleviate the disturbances resulting from noisy sensor data and variations in operating environment?

In the 1980s, Marc Raibert and associates at the MIT Leg Laboratory developed a tethered but self-balancing 3D one-leg hopping robot. For the current project, students would simulate the hopper dynamics, then determine feasibility, design, and build an autonomous (untethered) robot, using modern sensors and actuators.

Hobbyists worldwide are programming simple, low-cost biped robots with joints driven by RC (radio-control) servo motors, and achieving a variety of impressive motions and behaviors. How well can our RC robot be made to walk, or perform other related behaviors? This needs some mechanical design and electronics, but mostly mechanical analysis/simulation and C (or potentially C++) programming.

The motion of Steinkamp's new hopping toy is both remarkable and improbable - how does it do that? Develop a computer simulation of this device.

For an individual muscle, what is the relationship between force, speed, perceived effort, and energy expenditure? This could involve computer modeling, muscle physiology, test fixture construction, and human subject testing

Build and test a machine for on-land crew training that better simulates on-water rowing than any ergometer now available.

If cyclists’ legs and feet are constrained to follow a predetermined path during pedaling, will their power output or efficiency be improved? Student(s) would perform computer modeling, construction of an experimental stationary bicycle, power and VO2 testing.

Analyze the biomechanics of rowing with computer models.

Does the smoother pedaling motion of a bicycle moving quickly in high gear permit higher power output? Build an experimental stationary bicycle, and then work with human subjects to measure power.

Can we set a new world record in short-term human power output, using a machine that draws efficiently on most of the major muscles of the body? This would involve mechanical design, sensors, controls, and human-subject testing.

Could people run faster backwards than they can forwards under low gravity? Some theory points that way. This project would involve simple biomechanics experiments.

The Cornell gait lab has extensive facilities for gait analysis and motion capture. Learn to use this technology to analyze robot and human locomotion.

How can you sail downwind faster than the wind speed? Search the literature, explore the theory, and perhaps build a model. The semester report will be an article submitted for publication.

If you drop an apple and a vertical chain simultaneously, from a position with the apple level with the top link of the chain, the top link of the chain can hit the floor first. How does this work? Construct models (computer and physical). Demonstrate and publish the results.

How does an ellipsoid roll down a ramp? What happens when it becomes airborne and bounces? Prospects for a scientific toy are promising.