View Full Version : Cassie, bipedal robot, Dynamic Robotics Laboratory, Corvallis, Oregon, USA
Airicist
4th December 2016, 07:56
Developer - Dynamic Robotics Laboratory (https://pr.ai/showthread.php?10325)
Agility Robotics, Inc. (https://pr.ai/showthread.php?t=15982)
Principal Investigator: Dr. Jonathan Hurst (https://pr.ai/showthread.php?10326)
Cassie is the next generation of the recently retired ATRIAS (https://pr.ai/showthread.php?10327).
Airicist
4th December 2016, 09:19
Article "Oregon State builds human-like robot, launches 'next revolution' (https://www.oregonlive.com/teens/index.ssf/2016/06/oregon_state_builds_human-like.html)"
June 25, 2016
Airicist
1st April 2017, 06:19
https://youtu.be/Is4JZqhAy-M
Cassie - Next Generation Robot
Published on Feb 9, 2017
Engineers at Oregon State University are creating the next generation of bipedal robots.
Airicist
1st April 2017, 06:20
https://youtu.be/Co84Tb0x1fc
Don't run away, Cassie!
Published on Feb 9, 2017
Cassie gets a little excited while out for a walk.
Airicist
17th August 2017, 21:04
https://youtu.be/2DOw47x6bYo
Launch customer robots hanging out
Published on Aug 17, 2017
Time-lapse of all three launch customer robots dynamically balancing and demonstrating the "gaze" behavior that indicates power-on status.
Airicist
6th September 2017, 14:18
https://youtu.be/teCBAppDdL8
Cassie's arrival | Michigan Robotics
Published on Sep 6, 2017
Michigan Robotics’ latest member arrived last week in a black plastic case about two feet to a side, buried beneath a layer of foam cubes and crouched on a metal calibration stand.
The latest model, dubbed Cassie Blue by the Michigan team, has control over two more joints in each leg – motors for hip rotation and at the ankle for extra stability. Not only do these give Cassie the potential to be better at the independent walking pioneered by its predecessor, MARLO, but it opens a host of new possibilities.
Jessy Grizzle, the director of Michigan Robotics and the Elmer G. Gilbert Distinguished University Professor of Engineering, is especially interested in mounting a camera on Cassie and incorporating fast image processing on its extra chip, enabling the robot to use video to identify large changes in terrain visually.
This, Grizzle says, will prepare Cassie for Michigan’s “robot torture track” – the art installation known as the Wave Field. The field of three-foot earthen mounds totally destroyed one of MARLO’s knees last summer when the robot stepped blindly from the top of one hillock into the ditch.
“You’ve seen kids running freely over the Wave Field?” Grizzle asked. “That’s what this robot should be able to do.”
Airicist
6th September 2017, 20:35
https://youtu.be/DS18SeuMqtA
Cassies take a tour of Agility Robotics
Published on Sep 6, 2017
Two Cassies decide to take a walking tour of our office. No CG: 100% actual robots.
Airicist
30th October 2017, 12:02
https://youtu.be/YBVcHJdHoNs
Cassie Blue walking in the rain
Published on Oct 30, 2017
Cassie Blue is walking with control laws developed at Michigan.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI).
Airicist
31st October 2017, 01:17
https://youtu.be/bioubaR40C8
Cassie Blue's first outdoor adventure
Published on Oct 30, 2017
Don't miss the bloopers at the end!
Cassie Blue is walking with control laws developed at Michigan.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
31st October 2017, 01:18
https://youtu.be/h6hnfCo4a00
The original record of Cassie's first outdoor experiment
Published on Oct 30, 2017
Cassie Blue is walking with control laws developed at Michigan.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
16th November 2017, 22:45
https://youtu.be/onYv7mfHo9k
Who says Cassie Blue is afraid of the dark?
Published on Nov 16, 2017
Cassie Blue is walking with control laws developed at Michigan.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
8th December 2017, 10:44
https://youtu.be/ZL8ddCQFuXU
Toward the Robots of Science Fiction - A. Ames - 12/6/2017
Published on Dec 7, 2017
"Toward the Robots of Science Fiction, " by Aaron D. Ames, Bren Professor of Mechanical and Civil Engineering and Control and Dynamical Systems, Caltech
Science fiction has long promised a world of robotic possibilities: from humanoid robots in our everyday lives, to wearable robotic devices that restore and augment human capabilities, to swarms of autonomous robotic systems forming the backbone of the cities of the future, to robots enabling exploration of the cosmos.
Achieving the promise of science fiction will require imbuing machines with the dynamic locomotion behaviors that humans display with deceptive ease—navigating everything from daily environments to uneven and uncertain terrain with efficiency and robustness.
This talk will present the first steps toward achieving this goal on bipedal and humanoid robots with the result being dynamic and efficient locomotion displaying the hallmarks of natural human walking. The translation of these ideas to robotic assistive devices along with a wide range of safety-critical systems will be demonstrated with a view toward realizing the robots of science fiction.
Airicist
15th December 2017, 21:53
https://youtu.be/jhQMq6vpnAo
Cassie Blue Tours the World's #1 Ranked Dental School
Published on Dec 15, 2017
Cassie Blue is walking with control laws developed at Michigan in this publication: Da, X., Harib, O., Hartley, R., Griffin, B., & Grizzle, J. W. (2016). From 2D design of underactuated bipedal gaits to 3D implementation: Walking with speed tracking. IEEE Access, 4, 3469-3478.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
22nd March 2018, 22:21
https://youtu.be/f77NAz0HitI
Robot playdate
Published on Mar 22, 2018
Cassie had a meet-and-greet with a four-legged friend during one of our visits to Playground.
Airicist
8th April 2018, 20:06
https://youtu.be/ugEu0hzC8Xg
Robot on a Segway - a teasser of more to come
Published on Apr 8, 2018
Cassie Blue was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The robot has been operating Since October 2017 with control laws designed at Michigan and funded by NSF Grant NRI-1525006.
Airicist
24th April 2018, 21:10
https://youtu.be/cGb3bE6ZwrQ
Cassie Blue walks through fire
Published on Apr 24, 2018
The prescribed burn shown in the video took place on April 21, 2018 on the North Campus of the University of Michigan. The burns are done to promote the ecology of native species; you can learn more here "Controlled Burns (https://www.a2gov.org/departments/Parks-Recreation/NAP/Pages/PrescribedEcologicalBurns.aspx)"
Robots can be protected against flames and they do not suffer from smoke inhalation. While Cassie Blue has no arms to carry you out, coming generations of robots will be able to assist you.
Cassie Blue is walking with control laws developed at Michigan in this publication: Da, X., Harib, O., Hartley, R., Griffin, B., & Grizzle, J. W. (2016). From 2D design of underactuated bipedal gaits to 3D implementation: Walking with speed tracking. IEEE Access, 4, 3469-3478.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
11th May 2018, 13:38
https://youtu.be/0gauVSUJzd0
Cassie Blue рones рer Segway riding skills
Published on May 11, 2018
Cassie Blue is controlling the motion of the Segway by body lean, just as a human rider would do. To turn, she leans into the middle bar of the Segway with her "shin".
The feedback law being used is Cassie Blue's normal standing controller. Yukai Gong, via the remote controller, commands Cassie Blue's body lean. It would be much better if we modified the controller system so that we just sent desired speed and turn radius to the robot, and let feedback take care of the rest! Because we did not do that, we ran into a bit of trouble: the Segway has a built-in speed limiter. We'd be commanding Cassie Blue to go faster, faster, faster, by leaning further and further forward, and when the Segway would hit is speed limit, Cassie Blue would simply tip forward and fall of the Segway!
Airicist
11th May 2018, 13:39
https://youtu.be/VoD7hbssu-M
Cassie Blue plays in the sand
Published on May 11, 2018
Cassie Blue is operating with our normal flat-ground feedback controller. We made no changes to accommodate the sand. We were quite surprised that she could walk barefoot in such soft sand! Our best guess is that this ability is a fortunate outcome of the way we control her "feet". When a leg is in the air, we control the corresponding foot to be level with the ground. When a leg is in contact with the ground, we set the "ankle" torque to zero. In other words, we are treating the robot as being underactuated.
The Discovery Channel was filming on May 9, 2018.
Cassie Blue is walking with control laws developed at Michigan in this publication: Da, X., Harib, O., Hartley, R., Griffin, B., & Grizzle, J. W. (2016). From 2D design of underactuated bipedal gaits to 3D implementation: Walking with speed tracking. IEEE Access, 4, 3469-3478.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
24th May 2018, 19:00
https://youtu.be/rfrHctCH2zw
Cassie on stairs: forwards, backwards, and sideways
Published on May 24, 2018
Cassie shows off some new abilities, and - bonus! - avoids crashing through the nearby window.
Airicist
24th May 2018, 19:01
https://youtu.be/_6XBZHvt7bk
Cassie: walk in the park
Published on May 24, 2018
Practicing inside is all well and good, but nothing beats a walk outdoors on a spring day.
Airicist
24th May 2018, 19:03
https://youtu.be/svwQ4a5L1tk
Bipedal Robot in Wildfire Awareness Report, Action News Now, Chico-Redding, California
Published on May 24, 2018
Video posted with permission:
Spencer Joseph, Reporter
Lorraine Dechter, Assignment Desk/Producer
Action News Now, Chico-Redding, CA
Cassie Blue is walking with control laws developed at Michigan in this publication: Da, X., Harib, O., Hartley, R., Griffin, B., & Grizzle, J. W. (2016). From 2D design of underactuated bipedal gaits to 3D implementation: Walking with speed tracking. IEEE Access, 4, 3469-3478.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
24th May 2018, 21:38
https://youtu.be/XoY5vl3rQR8
Trolling Agility Robotics (or, We Love Your Bipedal Robot!)
Published on May 24, 2018
We have been working toward a principled approach to stair climbing when we saw Agility Robotics tweet on 25 May 2018 involving their Cassie walking on platforms. We could not resist posting what we have so far as well. Keep building those awesome robots!
Cassie Blue is walking with control laws developed at Michigan in this publication: Da, X., Harib, O., Hartley, R., Griffin, B., & Grizzle, J. W. (2016). From 2D design of underactuated bipedal gaits to 3D implementation: Walking with speed tracking. IEEE Access, 4, 3469-3478.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
25th May 2018, 12:05
https://youtu.be/oyKCBc9CZUA
Bipedal Robot Cassie Blue's Clutzy Operators and a Segway: Bloopers
Published on May 25, 2018
Operator Error #1: Gantry hit the sidewalk and stopped, pulling Cassie Blue off the Segway.
Operator Error #2: Bruce thought we could go through the wall in some quantum mechanics trick!
Controller Error #3: Bad interactions between the Segway's control loop and Cassie Blue's controller. That is called instability!
Controller Error #4: See above!
Operator Error #5: Yukai commanded too big of a lean!
Segway Error #6: The Segway hit its internal speed limit and leaned backward to slow down without asking permission first from Cassie Blue, the rider. That was a really bad idea!
Cassie Blue is operating with control laws developed at Michigan in this publication: Da, X., Harib, O., Hartley, R., Griffin, B., & Grizzle, J. W. (2016). From 2D design of underactuated bipedal gaits to 3D implementation: Walking with speed tracking. IEEE Access, 4, 3469-3478.
Cassie was built by Agility Robotics. The robot's purchase was enabled by funding from NSF Inspire Grant ECCS-1343720 and Toyota Research Institute (TRI). The work on the control law was funded by NSF Grant NRI-1525006.
Airicist
15th June 2018, 15:21
https://youtu.be/av8xjJYQsvE
Cassie vs scattered boards
Published on Jun 15, 2018
Cassie is an efficient, compliant, dynamic bipedal robot. Here in the Dynamic Robotics Lab directed by Dr. Jonathan Hurst we are researching control strategies that exploit the robot's natural dynamics. The controller here is demonstrating robust blind walking over unperceived disturbances.
Funded by
DARPA Award WN911NF-16-1-0002
National Science Foundation Graduate Research Fellowship Grant No. 1314109-DGE
Thanks to Agility Robotics for designing and producing Cassie.
Airicist
20th September 2018, 11:44
https://youtu.be/UhXly-5tEkc
Feedback control of a Cassie bipedal robot: walking, standing, and riding a Segway
Published on Sep 20, 2018
Airicist
30th January 2019, 23:21
https://youtu.be/oGxsbborgQg
Bipedal Robot Cassie Blue and Maize vs the Polar Vortex!
Published on Jan 30, 2019
30 January 2019: We wanted to see how long Cassie could operate during the "polar vortex" that has shutdown the campus! Well, she did one hour and two minutes of continuous operation at -22 C (-8 F).
Airicist
20th February 2019, 22:01
https://youtu.be/qV-92Bq96Co
Cassie: dynamic planning on stairs
Published on Feb 20, 2019
This video shows the progression of increasingly complex gait strategies from ATRIAS through recent results with Cassie. Specifically, Cassie's controller now includes planned footstep placements in addition to dynamic balancing, allowing access to substantially more complicated terrains.
Airicist
24th April 2019, 01:24
https://youtu.be/NYooAyAC0kA
Cassie: dynamic walking with compliance
Published on Apr 23, 2019
Dynamic walking with Cassie, from simulation to outdoor experiments, utilizing the full dynamics of the robot to generate the walking gait. This includes leverage the compliance, i.e., springs, to achieve the desired walking gaits.
Airicist
11th May 2019, 19:52
https://youtu.be/TgFrcrARao0
Iterative gait design for Cassie (https://arxiv.org/abs/1903.09537)
Published on Apr 26, 2019
Results of using DeepRL on a Cassie robot. Policies are trained in simulation and directly applied on hardware.
Airicist
4th July 2019, 11:04
https://youtu.be/f-FvcHOQXPc
Cassie vs Spin: a true story
Published on Jul 4, 2019
Cassie Blue gets her independence on July 4th! In this video, Cassie Blue is navigating autonomously. Right now, her world is very small, the Wavefield at the University of Michigan, where she is told to turn left at any intersection. You're right, that is not a lot of independence, but it's a first step away from a human and an RC controller!
Using a RealSense RGBD Camera, Cassie Blue is building a 3D semantic map in real time that identifies sidewalks, grass, poles, bicycles, and buildings. From the semantic map, occupancy and cost maps are built with the sidewalk identified as walk-able area and everything else considered as an obstacle. A planner then sets a goal to stay approximately 50 cm to the left of the sidewalk's edge and plans path around obstacles and corners. The path is translated into waypoints that are achieved via Cassie Blue's gait controller.
While there is a lot left to do, we knew that Cassie Blue would enjoy celebrating Independence Day with you!
Airicist
11th July 2019, 20:53
https://youtu.be/CBdDcsb1DWM
Published on Jul 11, 2019
Cassie (30 Kg) is walking at a comfortable speed for human while carrying her (9 Kg) perception system. The original controller is used, with the definition of the virtual leg modified to be the line segment connecting overall Center of Mass to her toe. This moves the "control point" up about 10 cm.
Airicist
2nd April 2020, 21:00
https://youtu.be/0nM7lml7kiE
Learning spring mass locomotion on Cassie
Apr 2, 2020
This is the video that accompanies our submission to the 2020 RSS conference which is currently under review. Once the review process is complete, we will post our full text preprint on an open access site and add a link here. The full title is:
Learning Spring Mass Locomotion: Guiding Policies with a Reduced Order Model
Kevin Green, Yesh Godse, Jeremy Dao, Ross L. Hatton, Alan Fern and Jonathan Hurst
Airicist
24th June 2020, 21:59
https://youtu.be/fMKVB822dOw
Cassie Blue doing her thing on the U. of Michigan North Campus lawn
Jun 24, 2020
Airicist
25th June 2020, 15:22
https://youtu.be/nRYXianfoQk
Cassie vs the hill
Jun 25, 2020
The controller is designed for flat ground and treats the 22 degree slope as a disturbance. As you can "see'', Cassie is not equipped with a vision system in this video. This is called "blind walking", though, to be completely fair, the operator does have vision and it's hard to know if he is making subtle corrections over the RC controller. He says not....
Airicist
2nd July 2020, 20:59
https://youtu.be/-35xJfFMDkE
Cassie Blue goes a ridge too far
Jul 2, 2020
As our control skills increase, we are more and more impressed by what a Cassie bipedal robot can do. Those who have been following our channel, know that we always shows the limitations of our work. So while there is still much to do, you gotta like the direction things are going. Later this year, you will see this controller integrated with our real-time planner and perception system. Autonomy with agility! Watch out for us!
Valley 1: 0:00
Valley 2: 1:11
Crossing Ridges 1: 2:03
Crossing Ridges 2: 2:43
Airicist
21st July 2020, 21:16
https://youtu.be/lu_jcL84wNE
Bipedal robot Cassie Blues reaches a walking speed of 2.1m/s
Jul 21, 2020
We are using distance traveled divided by time to estimate the velocity. The distance traveled is marked at time 0:39 of the video; see also here (https://youtu.be/lu_jcL84wNE?t=39). The plot shown at time 0.37 (https://youtu.be/lu_jcL84wNE) is an independent estimate from a Kalman Filter. When we get a chance, we'll try the same experiment on a running track without the gantry.
Can Cassie walk faster? Probably not, without running!
Airicist
25th October 2020, 21:18
https://youtu.be/Wb0tIWBrjmc
Cassie robot learns to hop, run and skip
Oct 25, 2020
These are some preliminary results of our lab's new work on using reinforcement learning to train neural networks to imitate common bipedal gait behaviors, without using any motion capture data or reference trajectories. Our method is described in an upcoming submission to ICRA 2021. Work by Jonah Siekmann and Yesh Godse.
Airicist
10th April 2021, 03:10
Article "Forget Boston Dynamics. This robot taught itself to walk (https://www.technologyreview.com/2021/04/08/1022176/boston-dynamics-cassie-robot-walk-reinforcement-learning-ai)"
Slick, viral videos from Boston Dynamics are impressive but teaching a robot to walk by itself is a lot harder.
by Will Douglas (https://www.linkedin.com/in/will-douglas-heaven-843358b)
April 8, 2021
Airicist
19th May 2021, 03:10
https://youtu.be/MPhEmC6b6XU
Robot learns to climb stairs blind
May 19, 2021
We successfully used reinforcement learning to train a recurrent neural network to control our bipedal robot Cassie to climb stairs without any perception sensors such as LIDAR or cameras.
This video accompanies our submission to the 2021 Robotics: Science and System conference. A preprint of the full paper can be found here https://arxiv.org/abs/2105.08328.
"Blind Bipedal Stair Traversal via Sim-to-Real Reinforcement Learning" Jonah Siekmann, Kevin Green, John Warila, Alan Fern, Jonathan Hurst
Airicist
19th May 2021, 03:11
https://youtu.be/nuhHiKEtaZQ
Reliability tests of a learned stair climbing controller
May 19, 2021
To show how reliable our learned controller is, we wanted to publish an uninterrupted video of ten stair ascents and ten stair descents. The controller being tested is described in our 2021 Robotics: Science and Systems paper "Blind Bipedal Stair Traversal via Sim-to-Real Reinforcement Learning" Preprint: https://arxiv.org/abs/2105.08328
Airicist
23rd May 2021, 09:07
Article "Agility Robotics' Cassie Is Now Astonishingly Good at Stairs (https://spectrum.ieee.org/automaton/robotics/humanoids/agility-robotics-cassie-stairs)"
Using only proprioceptive sensors, this bipedal robot is probably better at stairs than you are
by Evan Ackerman (https://www.linkedin.com/in/evan-ackerman-31112216)
May 20, 2021
Airicist
14th July 2021, 19:04
https://youtu.be/gE3Y-2Q3gco
Fully autonomy on the Wave Field 2021
Jun 4, 2021
During the dark of night, using LiDAR for eyes, Cassie Blue is operating fully autonomously on the University of Michigan Wave Field. The terrain is challenging and was not pre-mapped.
The LiDAR and IMU data are fused in real-time to form an elevation map. A CLF- RRT* planner is running at 5Hz with reactive re-planning performed at 300 Hz on the basis of a Control Lyapunov Function (CLF). The reactive planner provides velocity commands to a One-step Ahead Gait Controller based on Angular Momentum https://youtu.be/V36DCsc6iio . The paper https://arxiv.org/abs/2105.08170 provides full details on the controller. The planner is currently unpublished.
This work was presented at the 5th ICRA Workshop on Legged Robotics on 4 June 2021.
docs.google.com/presentation/d/1onC0VqRHnzqb3jgufmPRgeIwla74jQLFmeLGrjASdRw/edit#slide=id.g4cee01c7e5_0_24 (https://docs.google.com/presentation/d/1onC0VqRHnzqb3jgufmPRgeIwla74jQLFmeLGrjASdRw/edit#slide=id.g4cee01c7e5_0_24)
Airicist
28th July 2021, 15:29
https://youtu.be/FSaSjd_HOaI
Jul 27, 2021
To encourage us to push the limits of reliability, energy efficiency and speed we attempted to run a 5k (5 kilometers, 3.1 miles) with our bipedal robot Cassie.
Our final goal was to run a 5k through the Oregon State University campus. That can be seen here https://youtu.be/dY57qnD_O7U
This turf 5k was a test before we attempted the campus run. We finished with a time of 43:59.53 which means we had an average speed of 1.89 m/s. This is the fastest bipedal robot 5k at the time of posting.
https://youtu.be/dY57qnD_O7U
Machine learning breakthrough: Robot runs a 5k
Jul 27, 2021
The OSU Dynamic Robotics Laboratory's research team, led by Agility Robotics’ Jonathan Hurst, combined expertise from biomechanics and robot controls with new machine learning tools to accomplish something new: train a bipedal robot to run a full 5K on a single battery charge! This industry-first invention will unleash new levels of robot performance. Today, some of these same researchers are Agility Robotics employees busily applying their know-how to Digit, so just wait to see what we have in store.
"Cassie the bipedal robot runs a 5K (https://techcrunch.com/2021/07/27/cassie-the-bipedal-robot-runs-a-5k)"
by Brian Heater (https://www.linkedin.com/in/brian-heater-094a921)
July 28, 2021
Airicist2
7th November 2021, 11:23
https://youtu.be/utWqXZwTIbQ
Terrain-Aware Foot Placement for Bipedal Locomotion Combining MPC, Virtual Constraints, and the ALIP
Nov 4, 2021
Title: Terrain-Aware Foot Placement for Bipedal Locomotion Combining Model Predictive Control, Virtual Constraints, and the ALIP
Authors: Grant Gibson, Oluwami Dosunmu-Ogunbi, Yukai Gong, Jessy Grizzle
Preprint link: https://arxiv.org/abs/2109.14862
Abstract:
This paper draws upon three themes in the bipedal control literature to achieve highly agile, terrain-aware locomotion. By terrain aware, we mean the robot can use information on terrain slope and friction cone as supplied by state-of-the-art mapping and trajectory planning algorithms. The process starts with abstracting from the full dynamics of a Cassie 3D bipedal robot, an exact low-dimensional representation of its centroidal dynamics, parameterized by angular momentum. Under a piecewise planar terrain assumption, and the elimination of terms for the angular momentum about the robot's center of mass, the centroidal dynamics become linear and has dimension four. Four-step-horizon model predictive control (MPC) of the centroidal dynamics provides step-to-step foot placement commands. Importantly, we also include the intra-step dynamics at 10 ms intervals so that realistic terrain-aware constraints on robot's evolution can be imposed in the MPC formulation. The output of the MPC is directly implemented on Cassie through the method of virtual constraints. In experiments, we validate the performance of our control strategy for the robot on inclined and stationary terrain, both indoors on a treadmill and outdoors on a hill.
We thank Margaret Eva Mungai, Jennifer Humanchuk, and Jianyang Tang for assistance in experiments.
Airicist2
21st November 2021, 23:41
https://youtu.be/3HVJotA-w4Y
Cassie autonomously navigates around obstacles
Nov 17, 2021
Cassie Blue navigates around furniture treated as obstacles in the atrium of the Ford Robotics Building at the University of Michigan.
Airicist2
13th December 2021, 17:09
https://youtu.be/PT2mVaKTdT8
Cassie autonomously navigates in four long corridors (200 meters)
Nov 29, 2021
Cassie Blue autonomously navigates on the second floor of the Ford Robotics Building at the University of Michigan. The total traverse distance is 200 m (656.168 feet).
The LiDAR and IMU data are fused in real-time to form an elevation map. The system consists of a low-frequency planning thread (5 Hz) to find an asymptotically optimal path and a high-frequency reactive thread (300 Hz) to accommodate robot deviation. The planning thread includes: a multi-layer local map to compute traversability for the robot on the terrain; an anytime omnidirectional Control Lyapunov Function (CLF) for use with a Rapidly Exploring Random Tree Star (RRT*) that generates a vector field for specifying motion between nodes; a sub-goal finder when the final goal is outside of the current map; and a finite-state machine to handle high-level mission decisions. The paper [ https://arxiv.org/abs/2108.06699 ] provides full details on the reactive planning system.
Airicist2
5th August 2022, 23:09
https://youtu.be/AmPvQMpIHSw
Terrain - Adaptive, ALIP-based bipedal locomotion controller via MPC and virtual constraints - extended
Jul 30, 2022
Extended Video for IROS 2022 Accepted Paper (https://arxiv.org/abs/2109.14862)
Authors: Grant Gibson, Oluwami Dosunmu-Ogunbi, Yukai Gong, and Jessy Grizzle
Title: Terrain-Adaptive, ALIP-Based Bipedal Locomotion Controller via Model Predictive Control and Virtual Constraints
Abstract:
This paper presents a gait controller for bipedal robots to achieve highly agile walking over various terrains given local slope and friction cone information. Without these considerations, untimely impacts can cause a robot to trip and inadequate tangential reaction forces at the stance foot can cause slippages. We address these challenges by combining, in a novel manner, a model based on an Angular Momentum Linear Inverted Pendulum (ALIP) and a Model Predictive Control (MPC) foot placement planner that is executed by the method of virtual constraints. The process starts with abstracting from the full dynamics of a Cassie 3D bipedal robot, an exact low-dimensional representation of its center of mass dynamics, parameterized by angular momentum. Under a piecewise planar terrain assumption and the elimination of terms for the angular momentum about the robot's center of mass, the centroidal dynamics about the contact point become linear and have dimension four. Importantly, we include the intra-step dynamics at uniformly-spaced intervals in the MPC formulation so that realistic workspace constraints on the robot's evolution can be imposed from step-to-step. The output of the low-dimensional MPC controller is directly implemented on a high-dimensional Cassie robot through the method of virtual constraints. In experiments, we validate the performance of our control strategy for the robot on a variety of surfaces with varied inclinations and textures.
Airicist2
1st October 2022, 01:51
https://youtu.be/DdojWYOK0Nc
Cassie sets world record for 100M run (https://today.oregonstate.edu/news/bipedal-robot-developed-oregon-state-achieves-guinness-world-record-100-meters)
Sep 27, 2022
Cassie, the robot, clocked the historic time of 24.73 seconds at the Whyte Track and Field Center, starting from a standing position and returning to that position after the sprint, with no falls. This incredible achievement was accomplished through robot learning and almost a year of simulation, condensed down to a matter of weeks.
Cassie was invented at the Oregon State University College of Engineering and produced by Agility Robotics. Through this work Cassie has established a Guinness World Record for the fastest 100 meters by a bipedal robot.
Airicist2
25th November 2023, 16:27
https://youtu.be/utANK8jTwuI?si=g1cnFiquzayhTFxk
Cassie on a Moving Walkway
Nov 21, 2023
Airicist2
7th April 2024, 17:23
https://youtu.be/pKNiDnennBM?si=43-UICQG3p_1UyMp
Cassie walks on sand, gravel, and rocks in the Robot Playground
Apr 1, 2024
Using the controller from the paper below, Cassie is able to walk on sand, gravel, and rocks inside the Robot Playground at the University of Michigan
Paper: https://arxiv.org/abs/2403.02486
Airicist2
12th April 2024, 21:34
https://youtu.be/6kgB8PqvLYU?si=cDu8qiB3cg29Q-XJ
Learning vision-based bipedal locomotion for challenging terrain
Apr 6, 2024
Supplementary video for 2024 IEEE International Conference on Robotics and Automation
arXiv: https://arxiv.org/abs/2309.14594
Abstract - Reinforcement learning (RL) for bipedal locomotion has recently demonstrated robust gaits over moderate terrains using only proprioceptive sensing. However, such blind controllers will fail in environments where robots must anticipate and adapt to local terrain, which requires visual perception. In this paper, we propose a fully-learned system that allows bipedal robots to react to local terrain while maintaining commanded travel speed and direction. Our approach first trains a controller in simulation using a heightmap expressed in the robot's local frame. Next, data is collected in simulation to train a heightmap predictor, whose input is the history of depth images and robot states. We demonstrate that with appropriate domain randomization, this approach allows for successful sim-to-real transfer with no explicit pose estimation and no fine-tuning using real-world data. To the best of our knowledge, this is the first example of sim-to-real learning for vision-based bipedal locomotion over challenging terrains.