Robomechanics Lab
Robomechanics Lab
  • 48
  • 105 731
Robust and Adaptive Rough Terrain Navigation Through Training in Varied Simulated Dynamics
We propose a model-based reinforcement learning approach for robust and adaptive long-horizon navigation in rough terrain environments. Offline, we train an adaptive dynamics model using a wide range of simulated systems. This model can adapt to any new system using state-transition observations from that system. Predictions from the model capture uncertainty about the system's exact dynamics stemming from insufficient observations. Online, we use a divergence constrained path planner to find routes that are robust to the robot's current understanding of dynamics. In our results, we show this allows for long-horizon driving strategies that are conservative when state-transition observations are limited but have improved performance after giving few state-transition observations.
Sean J. Wang, Si Heng Teng, and Aaron M. Johnson. “Robust and adaptive rough terrain navigation through training in varied simulated dynamics,” in ICRA Workshop on Resilient Off-road Autonomy, 2024
PDF: openreview.net/pdf?id=CAbuIrSY5m
zhlédnutí: 125

Video

Pay Attention to How You Drive: Safe and Adaptive Model-Based RL for Off-Road Driving
zhlédnutí 176Před 2 měsíci
Autonomous off-road driving is challenging as risky actions taken by the robot may lead to catastrophic damage. As such, developing controllers in simulation is often desirable as it provides a safer and more economical alternative. However, accurately modeling robot dynamics is difficult due to the complex robot dynamics and terrain interactions in unstructured environments. Domain randomizati...
Convergent iLQR for Safe Trajectory Planning and Control of Legged Robots - ICRA 2024
zhlédnutí 281Před 3 měsíci
In order to perform highly dynamic and agile maneuvers, legged robots typically spend time in underactuated domains (e.g. with feet off the ground) where the system has limited command of its acceleration and a constrained amount of time before transitioning to a new domain (e.g. foot touchdown). Meanwhile, these transitions can instantaneously change the system's state, possibly causing pertur...
LORIS: A Lightweight Free-Climbing Robot for Extreme Terrain Exploration - Presentation
zhlédnutí 354Před 3 měsíci
"LORIS: A Lightweight Free-Climbing Robot for Extreme Terrain Exploration," Paul Nadan, Spencer Backus, and Aaron M. Johnson, in IEEE International Conference on Robotics and Automation, May, 2024. CMU Robotics Institute: www.ri.cmu.edu/publications/loris-a-lightweight-free-climbing-robot-for-extreme-terrain-exploration/ Climbing robots can investigate scientifically valuable sites that convent...
Conflict-based Model Predictive Control for Scalable Multi-Robot Motion Planning
zhlédnutí 232Před 3 měsíci
This work presents a scalable and dynamically-feasible approach for generating motion plans for multiple robots simultaneously.
LORIS: A Lightweight Free-Climbing Robot for Extreme Terrain Exploration
zhlédnutí 16KPřed 3 měsíci
"LORIS: A Lightweight Free-Climbing Robot for Extreme Terrain Exploration," Paul Nadan, Spencer Backus, and Aaron M. Johnson, in IEEE International Conference on Robotics and Automation, May, 2024. CMU Robotics Institute: www.ri.cmu.edu/publications/loris-a-lightweight-free-climbing-robot-for-extreme-terrain-exploration/ Climbing robots can investigate scientifically valuable sites that convent...
The Simplest Walking Robot: A bipedal robot with one actuator and two rigid bodies
zhlédnutí 1,3KPřed 7 měsíci
The Simplest Walking Robot: A bipedal robot with one actuator and two rigid bodies
Hybrid iLQR Model Predictive Control for Contact Implicit Stabilization on Legged Robots
zhlédnutí 456Před 8 měsíci
Model predictive control (MPC) is a popular strategy for controlling robots but is difficult for systems with contact due to the complex nature of hybrid dynamics. To implement MPC for systems with contact, dynamic models are often simplified or contact sequences fixed in time in order to plan trajectories efficiently. In this work, we propose the hybrid iterative linear quadratic regulator (iL...
Grounding Robot Navigation in Self-Defense Law
zhlédnutí 1,3KPřed 9 měsíci
Presented at RO-MAN 2023
Proprioception and reaction for walking among entanglements
zhlédnutí 3,9KPřed 10 měsíci
Legged robots get snagged on entanglements like vines and cords. Can they detect these entanglement and quickly get unstuck? This is the talk presented by Justin Yim for our paper at the 2023 International Conference on Intelligent Robots and Systems: Justin K. Yim, Jiming Ren, David Ologan, Selvin Garcia Gonzalez, and Aaron M. Johnson "Proprioception and reaction for walking among entanglement...
Staged Contact Optimization: Combining Contact-Implicit and Multi-Phase Hybrid Trajectory Opt.
zhlédnutí 235Před 10 měsíci
Presentation for IROS 2023 paper: Michael R. Turski, Joseph Norby, and Aaron M. Johnson. "Staged Contact Optimization: Combining Contact-Implicit and Multi-Phase Hybrid Trajectory Optimization." In IEEE/RSJ International Conference on Intelligent Robots and Systems, October 2023 Trajectory optimization problems for legged robots are commonly formulated with fixed contact schedules. These multi-...
Proprioception and Tail Control Enable Extreme Terrain Traversal by Quadruped Robots - Presentation
zhlédnutí 192Před 10 měsíci
Presentation for IROS 2023 paper: Yanhao Yang, Joseph Norby, Justin K. Yim, and Aaron M. Johnson. "Proprioception and Tail Control Enable Extreme Terrain Traversal by Quadruped Robots." In IEEE/RSJ International Conference on Intelligent Robots and Systems, October 2023 Abstract: Legged robots leverage ground contacts and the reaction forces they provide to achieve agile locomotion. However, un...
Proprioception and Tail Control Enable Extreme Terrain Traversal by Quadruped Robots
zhlédnutí 2,4KPřed rokem
Legged robots leverage ground contacts and the reaction forces they provide to achieve agile locomotion. However, uncertainty coupled with contact discontinuities can lead to failure, especially in real-world environments with unexpected height variations such as rocky hills or curbs. To enable dynamic traversal of extreme terrain, this work introduces 1) a proprioception-based gait planner for...
Off-Nominal Rover Driving: Terrain Manipulation and Degraded Mobility Compensation
zhlédnutí 631Před rokem
Catherine Pavlov's thesis defense. Planetary exploration rovers are at the forefront of the exploration of our solar system, but their mobility in soft soils beyond nominal operation is poorly understood. A better understanding of wheel-soil interaction lets us do more with rovers in two key ways: 1) add manipulation capabilities through the use of nonprehensile manipulation, e.g. by using whee...
The Uncertainty Aware Salted Kalman Filter
zhlédnutí 275Před rokem
"The Uncertainty Aware Salted Kalman Filter: State Estimation for Hybrid Systems with Uncertain Guards," J. Joe Payne, Nathan J. Kong, and Aaron M. Johnson, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), October, 2022. arxiv.org/abs/2202.12729 We present a method for updating robotic state belief through contact with uncertain surfaces and apply this update to a ...
Microspine Design for Additive Manufacturing
zhlédnutí 522Před rokem
Microspine Design for Additive Manufacturing
Increasing the reliability of legged robots in the presence of uncertainty
zhlédnutí 827Před rokem
Increasing the reliability of legged robots in the presence of uncertainty
Quad-SDK
zhlédnutí 4KPřed 2 lety
Quad-SDK
Panel Discussion 2 - RSS 2022: The Science of Bumping Into Things
zhlédnutí 162Před 2 lety
Panel Discussion 2 - RSS 2022: The Science of Bumping Into Things
Pulkit Agrawal - RSS 2022: The Science of Bumping Into Things
zhlédnutí 397Před 2 lety
Pulkit Agrawal - RSS 2022: The Science of Bumping Into Things
Zachary Manchester - RSS 2022: The Science of Bumping Into Things
zhlédnutí 486Před 2 lety
Zachary Manchester - RSS 2022: The Science of Bumping Into Things
Alessandro Saccon - RSS 2022: The Science of Bumping Into Things
zhlédnutí 332Před 2 lety
Alessandro Saccon - RSS 2022: The Science of Bumping Into Things
Yan Gu - RSS 2022: The Science of Bumping Into Things
zhlédnutí 332Před 2 lety
Yan Gu - RSS 2022: The Science of Bumping Into Things
Yuki Shirai - RSS 2022: The Science of Bumping Into Things
zhlédnutí 123Před 2 lety
Yuki Shirai - RSS 2022: The Science of Bumping Into Things
Taylor Howell and Simon Le Cleac'h - RSS 2022: The Science of Bumping Into Things
zhlédnutí 245Před 2 lety
Taylor Howell and Simon Le Cleac'h - RSS 2022: The Science of Bumping Into Things
Panel Discussion 1 - RSS 2022: The Science of Bumping Into Things
zhlédnutí 93Před 2 lety
Panel Discussion 1 - RSS 2022: The Science of Bumping Into Things
Kevin Lynch - RSS 2022: The Science of Bumping Into Things
zhlédnutí 186Před 2 lety
Kevin Lynch - RSS 2022: The Science of Bumping Into Things
Huanbo Sun - RSS 2022: The Science of Bumping Into Things
zhlédnutí 107Před 2 lety
Huanbo Sun - RSS 2022: The Science of Bumping Into Things
Ken Goldberg - RSS 2022: The Science of Bumping Into Things
zhlédnutí 190Před 2 lety
Ken Goldberg - RSS 2022: The Science of Bumping Into Things
Maria Bauza - RSS 2022: The Science of Bumping Into Things
zhlédnutí 271Před 2 lety
Maria Bauza - RSS 2022: The Science of Bumping Into Things

Komentáře

  • @QXZ9027MKII
    @QXZ9027MKII Před 4 dny

    Top notch, What’s the purpose of the tail.

  • @rajehkumarmishra
    @rajehkumarmishra Před měsícem

    love this

  • @ProphetYoda
    @ProphetYoda Před 2 měsíci

    Miguel O’Hara

  • @saeshwaran
    @saeshwaran Před 2 měsíci

    what application did you use to simulate this environment?

  • @AnarcoPetista
    @AnarcoPetista Před 2 měsíci

    NASA needs to use robots on Mars, there is more correct soil than this to do a real test. Are we going to wait until 2040 to see robots on Mars or the Moon? unnecessary delay

  • @vingo9119
    @vingo9119 Před 2 měsíci

    So it can walk on the roof then correct?

  • @Beyondlife-x9x
    @Beyondlife-x9x Před 2 měsíci

    It's eazy hardware wut about it's software the puzzle By matlab !!! 😂😂😂

  • @Awesomes007
    @Awesomes007 Před 3 měsíci

    Love it!

  • @whitneydesignlabs8738
    @whitneydesignlabs8738 Před 5 měsíci

    Thanks for sharing, I was thinking about using a tail on a small quadruped robot, mostly to hold a camera so it can see its own feet position. But I think this tail could be multipurpose if I gave it 2 DoF. It could both be a balance aid, and a camera mount.

  •  Před 7 měsíci

    Good job...

  • @noelltaravati4177
    @noelltaravati4177 Před 9 měsíci

    Bravo! 😊

  • @papaveneti7759
    @papaveneti7759 Před rokem

    1. The red line is the CoM trajectory (output from global planner), the green line is the state trajectory and the arrows are the contact force trajectories F(t) (output from local body planner) and the 4 lines(green, red, yellow, blue) are the swing leg trajectories (output from Local Footstep planner)? 2. The foothold positions are calculated in the local footstep planner. But are required in the local body planner. So these two modules run multiple times to update their solutions based on the outcome of the other module?

  • @diegoalonsozamoracamacho6688

    What motors did you use?

  • @azarovwraith6352
    @azarovwraith6352 Před rokem

    This is amazing!!

  • @zhexu2908
    @zhexu2908 Před rokem

    Hi,Really impressive work. I am curious about the NMPC controller, how can it be really efficient with IPOPT solver? Does it apply the VBL-MPC (Representation-free model predictive control for dynamic motions in quadrupeds) proposed by Doctor Ding?

  • @ArifRahmanMY
    @ArifRahmanMY Před rokem

    Wow lots of good work there! Well done Nathan!

  • @MrDadarobot
    @MrDadarobot Před rokem

    This is awesome talk, very nice visualization. Great job Nathan.

  • @abdullahcakan
    @abdullahcakan Před 2 lety

    👍

  • @kissenk8983
    @kissenk8983 Před 2 lety

    Amazing work

  •  Před 2 lety

    Thank you for sharing.

  • @williamhuang5329
    @williamhuang5329 Před 2 lety

    Hanzhen harmonic drive gaer , robot gear joint , over 30 years experience

  • @chenyao4791
    @chenyao4791 Před 2 lety

    Also I check your Quad-SDK now, it's a great work now, but there have many branch and I wonder whicn I can use for the contact detection and proprioceptive terrain estimation as your paper said hare?

  • @digital6278
    @digital6278 Před 2 lety

    cool, one looks like an axolotl

  • @chenyao4791
    @chenyao4791 Před 2 lety

    Hi, where can I find your paper about this video shown, I wonder how to get your interesting terrain estimation method? Thank you for your share.

    • @robomechanicslab
      @robomechanicslab Před 2 lety

      Sorry for the slow reply! This work is being presented this week at ICRA 2022. The base code is now released as Quad-SDK, open source full-stack quadruped robot controller: Extended Abstract: www.andrew.cmu.edu/user/amj1/papers/Quad_SDK_ICRA_Abstract.pdf Code: github.com/robomechanics/quad-sdk The proprioceptive terrain estimation is presented separately: Extended Abstract: www.andrew.cmu.edu/user/amj1/papers/Proprioception_and_Tail_Control_ICRA_WS.pdf Additional video: czcams.com/video/uH6T1ETzjhM/video.html Video description now updated with these pointers. Hope this helps!

    • @chenyao4791
      @chenyao4791 Před 2 lety

      @@robomechanicslabYes, I will try your Quad-SDK in unitree go1 in these days, thank you for your kindly share here!

  • @jorgearay3241
    @jorgearay3241 Před 2 lety

    Hahahaha. Was funny

  • @royale9985
    @royale9985 Před 2 lety

    I think Popping and the Bunny Hop made me guffaw the most haha. What a lovely lab tradition! Just curious, is that black quadruped an A1 from Unitree? If not what model is it :O?

  • @bytefloww
    @bytefloww Před 2 lety

    this is so cool bro! i wish i could make one

  • @Thetitaniumsteppermotor

    Question why lion king? R hex: yes

  • @UtkucanUyank
    @UtkucanUyank Před 4 lety

    Hahahahahahha

  • @Bouldest1
    @Bouldest1 Před 4 lety

    Nice work Catherine!

  • @Buy-n-large
    @Buy-n-large Před 5 lety

    this is amazing! i was looking through the guide on GitHub and the only problem for me would be that there isn't any control system (or code for one) posted there, i would really like to build one of these, but i'm afraid that if i did, i wouldn't be able to do anything with it, since i wouldn't know how to set it up with a remote control or something like that.

    • @robomechanicslab
      @robomechanicslab Před 5 lety

      Hi @BNL! Sorry for the late reply, but we added Bluetooth control instructions to our Github: github.com/robomechanics/MiniRHex/tree/master/Bluetooth. You can now pair the robot with a laptop/PC or with your mobile device! But, you will have to buy additional bluetooth modules to work with the MiniRHex's board. If you have any questions and/or suggestions, let us know!

  • @nitrovent
    @nitrovent Před 5 lety

    Really cool! Since I learned about the Boston Dynamic's RHex years ago, I wanted to build one myself and now I just finished a clone of your MiniRHex within days. Nice work!

  • @kernel78
    @kernel78 Před 5 lety

    Emotional

  • @JemMawson
    @JemMawson Před 5 lety

    This is great. Thank you for making it so entertaining. I'm cry laughing here. :D