Miscellaneous

Distinguished University Professorship 2015 Lecture Series presented by Elmer G. Gilbert; Distinguished University Professor of Engineering, Jerry W. and Carol L. Levin; Professor of Engineering and the College of Engineering at the University of Michigan

The fields of control and robotics are working hand-in-hand to development bipedal machines that can realize walking motions with the stability and agility of a human being. Dynamic models for bipeds are hybrid nonlinear systems, meaning they contain both continuous and discrete elements, with switching events that are spatially driven by changes in ground contact. This talk will show how nonlinear control methods are enhancing the ability to achieve highly dynamic locomotion. The presented experiments will primarily focus on our past work on 2D (planar) bipedal robots; a new 3D robot is being installed at Michigan and we will show some of the preliminary results.
Jessy W. Grizzle received the Ph.D. in electrical engineering from The University of Texas at Austin in 1983 and in 1984 held an NSF-NATO Postdoctoral Fellowship in Science in Paris, France. Since September 1987, he has been with The University of Michigan, Ann Arbor, where he is the Jerry and Carol Levin Professor of Engineering. He jointly holds sixteen patents dealing with emissions reduction in passenger vehicles through improved control system design. Professor Grizzle is a Fellow of the IEEE and of IFAC. He received the Paper of the Year Award from the IEEE Vehicular Technology Society in 1993, the George S. Axelby Award in 2002, the Control Systems Technology Award in 2003, and the Bode Lecture Prize in 2012. His work on bipedal locomotion has been the object of numerous plenary lectures and has been featured in The Economist, Wired Magazine, Discover Magazine, Scientific American, Popular Mechanics and several television programs.

Click to expand...

Sam Ambler biped robot drawing at a Cheshire Open Studios art event in Knutsford, 21/22 July 2012

Click to expand...

This was another student project using R+Motion V.1

Click to expand...

This video illustrates the use of Phase-Indexed Iterative Learning Control on an underactuated dynamic walking robot (a compass-gait walker) designed and built at the Australian Centre for Field Robotics. In particular, note the large reduction in tracking error when ILC is turned on. Full details on the algorithm and experiments be found here:
"Phase-indexed ILC for control of underactuated walking robots"

Click to expand...

We equipped a small bipedal running robot with tactile sensors to perform terrain classification using machine learning.

Click to expand...

In this video we show the performance of a new force control approach for our humanoid LOLA. By using an explicit contact model and integrated CoM dynamics, we are able to walk from an unexpected platform of 5.5 cm height.

Click to expand...

Researchers from the University of Michigan have created a model to quantitatively measure the unpredictability of the movement pattern of the jerboa, a bipedal desert rodent. Ram Vasudevan, an assistant professor in mechanical engineering, in partnership with Talia Moore, a research fellow in ecology and evolutionary biology, used Information Theory to measure the randomness or unpredictability of this highly evasive animal. The researchers believe this new model for unpredictable movement can be applied to bipedal robots as a way to engineer unpredictability in their gait.

Click to expand...

In spite of the pandemic, Professor Hod Lipson’s Robotics Studio persevered and even thrived— learning to work on global teams, to develop protocols for sharing blueprints and code, and to test, evaluate, and refine their designs remotely. Equipped with a 3D printer and a kit of electronics prototyping equipment, our students engineered bipedal robots that were conceptualized, fabricated, programmed, and endlessly iterated around the globe in bedrooms, kitchens, backyards, and any other makeshift laboratory you can imagine.

Click to expand...

A team of scientists at the Max Planck Institute for Intelligent Systems and the University of California, Irvine constructed a robot leg that, like its natural model, is very energy efficient. BirdBot benefits from a foot-leg coupling through a network of muscles and tendons that extends across multiple joints. In this way, BirdBot needs fewer motors than previous legged robots and could, theoretically, scale to large size. On March 16th, the researchers will publish their work in Science Robotics.

Click to expand...

Implementation of a Large-scale Biped Robot Using Serial-Parallel Hybrid Leg Mechanism

By Kevin G Gim and Joohyung Kim

This paper presents our implementation of a large-scale biped robot utilizing Hybrid Leg, a 6 DoF serial-parallel mechanism, having lightweight structure, high payload and large workspace. We set our design goal to make a biped robot taller than an average human height. By applying the Hybrid mechanism and design optimization, the robot was built with a height of 1.84m and a weight of 29.05kg. The implemented robot is able to be actuated by the servo motors used in the smaller humanoid robot. The mechanical design of the robot is explained in detail and kinematics analysis is conducted for analytical solutions. Through multi-body dynamics simulations, the proposed robot design and its performance are verified. In addition, the preliminary performance evaluations for the robot hardware are conducted for a squat experiment and in-place walking experiment.

Click to expand...

We present a sim-to-real learning-based approach for real-world humanoid locomotion. Our controller is a causal Transformer trained by autoregressive prediction of future actions from the history of observations and actions. We hypothesize that the observation-action history contains useful information about the world that a powerful Transformer model can use to adapt its behavior in-context, without updating its weights. We do not use state estimation, dynamics models, trajectory optimization, reference trajectories, or pre-computed gait libraries. Our controller is trained with large-scale model-free reinforcement learning on an ensemble of randomized environments in simulation and deployed to the real world in a zero-shot fashion. We evaluate our approach in high-fidelity simulation and successfully deploy it to the real robot as well. To the best of our knowledge, this is the first demonstration of a fully learning-based method for real-world full-sized humanoid locomotion.

Click to expand...

The ‘Art & Science’ of storytelling is the secret to how we amaze our guests and delivers memorable experiences.  Check out this video as Disney Parks, Experiences & Products Chairman Josh D’Amaro shares how storytelling techniques will build on our legacy of creativity and innovation for a world that can always use just a little more happiness.

00:00 Intro
0:32 Adventure to Distant Lands – Tinker Bell
04:17 Using robotics to create characters

Click to expand...

Our new walk is based on our 2022 walk. To ensure stability, we use a regulation to modify the allowed rotation speed of the support foot’s joints. Thus, the different leg parts will still execute the intended motion, but based on the center of mass and the measured rotation errors of the support foot, some leg parts are slowed down if needed.

Additionally, to handle more extreme cases at higher walking speeds, a neural network is used to predict future joint position measurements to calculate future position errors.

The robots are now able to handle more difficult situations. Also, as an unintended effect, the robots lift up on the tip of the supporting foot, just like humans do.

Click to expand...