FastMimic: Model-based Motion Imitation for Agile, Diverse and Generalizable Quadrupedal Locomotion

Li, Tianyu; Won, Jungdam; Ha, Sehoon; Rai, Akshara

doi:10.48550/arxiv.2109.13362

Cited by 4 publications

(4 citation statements)

References 37 publications

(66 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While NLP-solver-based MPC often only applies u * to the system, resulting in open-loop MPC, the DDP-based MPC addtionally applies K * to the system, resulting in feedback MPC. The feedback MPC can account for more policy lags and earn more robustness [9], [26]. Further, during replanning, most single-shooting NLP solvers use u * from the previous planning to warm start the current optimization.…”

Section: B Schemes Of Maintaining Fast Convergencementioning

confidence: 99%

Versatile Real-Time Motion Synthesis via Kino-Dynamic MPC with Hybrid-Systems DDP

Li¹,

Zhang²,

Yu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Specialized motions such as jumping are often achieved on quadruped robots by solving a trajectory optimization problem once and executing the trajectory using a tracking controller. This approach is in parallel with Model Predictive Control (MPC) strategies that commonly control regular gaits via online re-planning. In this work, we present a nonlinear MPC (NMPC) technique that unlocks on-the-fly re-planning of specialized motion skills and regular locomotion within a unified framework. The NMPC reasons about a hybrid kinodynamic model, and is solved using a variant of a constrained Differential Dynamic Programming (DDP) solver. The proposed NMPC enables the robot to perform a variety of agile skills like jumping, bounding, and trotting, and the rapid transition between these skills. We evaluated the proposed algorithm with three challenging motion sequences that combine multiple agile skills, on two quadruped platforms, Unitree A1, and MIT Mini Cheetah, showing its effectiveness and generality.

show abstract

Section: B Schemes Of Maintaining Fast Convergencementioning

confidence: 99%

Versatile Real-Time Motion Synthesis via Kino-Dynamic MPC with Hybrid-Systems DDP

Li¹,

Zhang²,

Yu³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Thus, we also experiment with another model-predictive control (MPC) dynamic controller from [38], which commands joint torques directly. This controller has been applied to real-world A1 robot [52,53] and shows better tracking of desired velocities for our test robots, as compared to the Raibert controller from [22]. However, MPC is prohibitively slow and cannot be used for training RL policies.…”

Section: Kinematic and Dynamic Control For Visual Navigationmentioning

confidence: 99%

Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation

Truong¹,

Rudolph²,

Yokoyama³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

If we want to train robots in simulation before deploying them in reality, it seems natural and almost self-evident to presume that reducing the sim2real gap involves creating simulators of increasing fidelity (since reality is what it is). We challenge this assumption and present a contrary hypothesis -sim2real transfer of robots may be improved with lower (not higher) fidelity simulation. We conduct a systematic large-scale evaluation of this hypothesis on the problem of visual navigation -in the real world, and on 2 different simulators (Habitat and iGibson) using 3 different robots (A1, AlienGo, Spot). Our results show that, contrary to expectation, adding fidelity does not help with learning; performance is poor due to slow simulation speed (preventing large-scale learning) and overfitting to inaccuracies in simulation physics. Instead, building simple models of the robot motion using real-world data can improve learning and generalization.

show abstract

“…However, while controllers behave well in idealized simulated environments, they often struggle when transferred to the real world, exhibiting infeasible motor-control behaviors due to the difference between simulation and real-world, which is often referred to as the reality gap. Some approaches propose to address the reality gap with conventional optimization methods such as MPC, allowing the policy to adjust on the real-robot [30,68,40,76]. On the other hand, others have investigated methods that leverage real-world data, such as learning on real robots [22,21,62], identifying system parameters [29], or adapting policy behaviors [53,83,36].…”

Section: Related Work a Legged Robot Controlmentioning

confidence: 99%

Human Motion Control of Quadrupedal Robots using Deep Reinforcement Learning

Kim¹,

Sorokin²,

Lee³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

A motion-based control interface promises flexible robot operations in dangerous environments by combining user intuitions with the robot's motor capabilities. However, designing a motion interface for non-humanoid robots, such as quadrupeds or hexapods, is not straightforward because different dynamics and control strategies govern their movements. We propose a novel motion control system that allows a human user to operate various motor tasks seamlessly on a quadrupedal robot. We first retarget the captured human motion into the corresponding robot motion with proper semantics using supervised learning and post-processing techniques. Then we apply the motion imitation learning with curriculum learning to develop a control policy that can track the given retargeted reference. We further improve the performance of both motion retargeting and motion imitation by training a set of experts. As we demonstrate, a user can execute various motor tasks using our system, including standing, sitting, tilting, manipulating, walking, and turning, on simulated and real quadrupeds. We also conduct a set of studies to analyze the performance gain induced by each component.(Video 1

show abstract

FastMimic: Model-based Motion Imitation for Agile, Diverse and Generalizable Quadrupedal Locomotion

Cited by 4 publications

References 37 publications

Versatile Real-Time Motion Synthesis via Kino-Dynamic MPC with Hybrid-Systems DDP

Versatile Real-Time Motion Synthesis via Kino-Dynamic MPC with Hybrid-Systems DDP

Rethinking Sim2Real: Lower Fidelity Simulation Leads to Higher Sim2Real Transfer in Navigation

Human Motion Control of Quadrupedal Robots using Deep Reinforcement Learning

Contact Info

Product

Resources

About