= MIT Cheetah robot
{c}
The <MIT Cheetah robot> is an electrically actuated <quadrupedal robot> research platform used to investigate dynamic legged locomotion and control. Specific demonstrations must be identified by version and experiment. In the primary MIT account of a 2015 running-jump demonstration, onboard <Lidar> estimated obstacle height and distance, software adjusted the approach strides, and a subsequent calculation selected the jumping action.<MIT cheetah robot lands the running jump by Jennifer Chu>[29 May 2015, See, run, jump]{ref}
The experiment connects perception to physical contact. Detecting an obstacle does not by itself establish a feasible jump: the machine must reach a suitable takeoff state and supply forces within its motor and ground-contact constraints. Computation is performed while the robot approaches the obstacle, so available decision time is also a physical constraint.
An original ballistic approximation illustrates one requirement. If the center of mass must rise $\Delta h$ after takeoff, with gravity the only vertical force during flight, its launch velocity must satisfy
$$
v_y\ge\sqrt{2g\Delta h}.
$$
For $\Delta h=0.20\,\mathrm m$, this is about $1.98\,\mathrm{m/s}$. If the vertical velocity starts at zero and that change is generated over $0.10\,\mathrm s$ of contact, vertical <impulse> balance gives
$$
\overline F_y=mg+\frac{mv_y}{\Delta t}\approx3.02\,mg.
$$
These values are illustrative requirements for the stated center-of-mass model, not measured MIT parameters. Foot clearance, body orientation, leg motion and landing cannot be inferred from this one vertical calculation.
The primary report describes tests on both a treadmill and an indoor track. The available approach distance differed, and the reported clearance success differed as well. This makes the observation conditional on apparatus and trial conditions rather than a universal performance figure.<MIT cheetah robot lands the running jump by Jennifer Chu>[Optimal is best, feasible is better; treadmill and track testing paragraphs]{ref}
The scientific value of the platform is that a proposed controller has to close the loop through actual motors, contacts and sensor estimates. Successful trials demonstrate a capability in the tested conditions; robustness to other surfaces, disturbances or obstacles requires additional experiments.
Back to article page
