133 Hits in 4.1 sec

Planning in Learned Latent Action Spaces for Generalizable Legged Locomotion [article]

Tianyu Li, Roberto Calandra, Deepak Pathak, Yuandong Tian, Franziska Meier, Akshara Rai
2021 arXiv   pre-print
Once this latent space is learned, we plan over continuous latent actions in a model-predictive control fashion, using a learned high-level dynamics model.  ...  Hierarchical learning has been successful at learning generalizable locomotion skills on walking robots in a sample-efficient manner.  ...  This motivates the importance of model-based planning in learned latent action spaces for generalization versus learning a model-free policy.  ... 
arXiv:2008.11867v5 fatcat:4hy5buncavbwrm3ygu43hc2r4y

Learning Actionable Representations with Goal-Conditioned Policies [article]

Dibya Ghosh, Abhishek Gupta, Sergey Levine
2019 arXiv   pre-print
Most prior work on representation learning has focused on generative approaches, learning representations that capture all underlying factors of variation in the observation space in a more disentangled  ...  In this paper, we instead aim to learn functionally salient representations: representations that are not necessarily complete in terms of capturing all factors of variation in the observation space, but  ...  We thank Pim de Haan, Aviv Tamar, Vitchyr Pong, and Ignasi Clavera for helpful insights and discussions.  ... 
arXiv:1811.07819v2 fatcat:pphhr45mt5dt5mfbkfcb7gwuj4

Real-Time Trajectory Adaptation for Quadrupedal Locomotion using Deep Reinforcement Learning

Siddhant Gangapurwala, Mathieu Geisert, Romeo Orsolino, Maurice Fallon, Ioannis Havoutis
2021 2021 IEEE International Conference on Robotics and Automation (ICRA)  
Additionally, in order to capture terrain information, we include a latent representation of the height maps in the observation space of the RL environment as a form of exteroceptive feedback.  ...  We train a policy using deep reinforcement learning (RL) to introduce additive deviations to a reference trajectory in order to generate a feedback-based trajectory tracking system for a quadrupedal robot  ...  These corrective controllers have been employed for manipulation tasks [19] and also for legged locomotion [20] .  ... 
doi:10.1109/icra48506.2021.9561639 fatcat:ohxlvuhgnrc47oy4tsygvq5dna

Learning nonparametric policies by imitation

D.B. Grimes, R.P.N. Rao
2008 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems  
The novel contribution of this work is a method for learning a nonparametric policy which generalizes a fixed action plan to operate over a continuous space of task variation.  ...  The robot uses inference in a graphical model to learn sensor-based dynamics and infer a stable plan from a teacher's demonstration of an action.  ...  In this example the learned policy maps a 3D state representation of two latent posture dimensions and a foot sensor based dynamics dimension to actions in the 2D latent posture space.  ... 
doi:10.1109/iros.2008.4650778 dblp:conf/iros/GrimesR08 fatcat:2kplfe3dkvf4je6ldjmnczmqxi

Multi-Agent Manipulation via Locomotion using Hierarchical Sim2Real [article]

Ofir Nachum, Michael Ahn, Hugo Ponte, Shixiang Gu, Vikash Kumar
2019 arXiv   pre-print
Our method hinges on the use of hierarchical sim2real -- a simulated environment is used to learn low-level goal-reaching skills, which are then used as the action space for a high-level RL controller,  ...  Manipulation and locomotion are closely related problems that are often studied in isolation.  ...  Acknowledgments We thank Chad Richards, Byron David, Matt Neiss, Krista Reymann, Ben Eysenbach, Sergey Levine, and the rest of Robotics at Google for helpful thoughts and discussions.  ... 
arXiv:1908.05224v2 fatcat:ywya6tymcfbp3ev7kdvvct72di

Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers [article]

Ruihan Yang, Minghao Zhang, Nicklas Hansen, Huazhe Xu, Xiaolong Wang
2022 arXiv   pre-print
In this paper, we introduce LocoTransformer, an end-to-end RL method that leverages both proprioceptive states and visual observations for locomotion control.  ...  While learning-based locomotion has made great advances using RL, most methods still rely on domain randomization for training blind agents that generalize to challenging terrains.  ...  Acknowledgement: This work was supported, in part, by gifts from Meta, Qualcomm, and TuSimple.  ... 
arXiv:2107.03996v3 fatcat:l7tjgxb3prgv7p5vju4k5se2im

Language2Pose: Natural Language Grounded Pose Forecasting [article]

Chaitanya Ahuja, Louis-Philippe Morency
2019 arXiv   pre-print
This joint embedding space is learned end-to-end using a curriculum learning approach which emphasizes shorter and easier sequences first before moving to longer and harder ones.  ...  Generating animations from natural language sentences finds its applications in a a number of domains such as movie script visualization, virtual human animation and, robot motion planning.  ...  Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of National Science Foundation or Oculus VR.  ... 
arXiv:1907.01108v2 fatcat:jdrxu46mwfcbnkd47tx3nvj3vm

HyperDynamics: Meta-Learning Object and Agent Dynamics with Hypernetworks [article]

Zhou Xian, Shamit Lal, Hsiao-Yu Tung, Emmanouil Antonios Platanios, Katerina Fragkiadaki
2021 arXiv   pre-print
It outperforms existing dynamics models in the literature that adapt to environment variations by learning dynamics over high dimensional visual observations, capturing the interactions of the agent in  ...  We test HyperDynamics on a set of object pushing and locomotion tasks.  ...  MODEL UNROLLING AND ACTION PLANNING Action-conditioned dynamics models can be unrolled forward in time for long-term planning and control tasks.  ... 
arXiv:2103.09439v1 fatcat:42kmfav7bfedba3phtlgcoh2xq

D2RL: Deep Dense Architectures in Reinforcement Learning [article]

Samarth Sinha, Homanga Bharadhwaj, Aravind Srinivas, Animesh Garg
2020 arXiv   pre-print
While improvements in deep learning architectures have played a crucial role in improving the state of supervised and unsupervised learning in computer vision and natural language processing, neural network  ...  architecture choices for reinforcement learning remain relatively under-explored.  ...  INTRODUCTION Deep Reinforcement Learning (DRL) is a general purpose framework for training goal-directed agents in high dimensional state and action spaces.  ... 
arXiv:2010.09163v2 fatcat:vt2x32wbnnekfadpbuvafsgdqe

SimGAN: Hybrid Simulator Identification for Domain Adaptation via Adversarial Reinforcement Learning [article]

Yifeng Jiang, Tingnan Zhang, Daniel Ho, Yunfei Bai, C. Karen Liu, Sergey Levine, Jie Tan
2021 arXiv   pre-print
Our hybrid simulator combines neural networks and traditional physics simulation to balance expressiveness and generalizability, and alleviates the need for a carefully selected parameter set in System  ...  We show that our approach outperforms multiple strong baselines on six robotic locomotion tasks for domain adaptation.  ...  In the future, we plan to address these limitations, and test our method for transferring locomotion policies from simulation to a real Laikago robot.  ... 
arXiv:2101.06005v2 fatcat:3upuanwppffitenrenybagq5qa

Causal Curiosity: RL Agents Discovering Self-supervised Experiments for Causal Representation Learning [article]

Sumedh A. Sontakke, Arash Mehrjou, Laurent Itti, Bernhard Schölkopf
2021 arXiv   pre-print
The learned behavior allows the agents to infer a binary quantized representation for the ground-truth causal factors in every environment.  ...  We introduce causal curiosity, a novel intrinsic reward, and show that it allows our agents to learn optimal sequences of actions and discover causal factors in the dynamics of the environment.  ...  Many thanks also to Alexander Neitz for sourcing of the CEM planning code.  ... 
arXiv:2010.03110v4 fatcat:fk4vsjibc5bi7iyhjk7wwe6izu

A Survey on Model-based Reinforcement Learning [article]

Fan-Ming Luo, Tian Xu, Hang Lai, Xiong-Hui Chen, Weinan Zhang, Yang Yu
2022 arXiv   pre-print
For non-tabular environments, there is always a generalization error between the learned environment model and the real environment.  ...  As such, it is of great importance to analyze the discrepancy between policy training in the environment model and that in the real environment, which in turn guides the algorithm design for better model  ...  For high-dimensional state space such as images, representation learning that learns informative latent state or action representation will much benefit the environment model building so as to improve  ... 
arXiv:2206.09328v1 fatcat:cox3e76nhnewjpbdzmqgmgijnq

2019 Index IEEE Robotics and Automation Letters Vol. 4

2019 IEEE Robotics and Automation Letters  
Guadarrama-Olvera, J.R., +, LRA Oct. 2019 4418-4423 Robot Motion Planning in Learned Latent Spaces.  ...  ., +, LRA Oct. 2019 4298-4305 Robot Motion Planning in Learned Latent Spaces. Ichter, B., +, LRA July 2019 2407-2414 Safe Navigation With Human Instructions in Complex Scenes.  ...  Permanent magnets Adaptive Dynamic Control for Magnetically Actuated Medical Robots.  ... 
doi:10.1109/lra.2019.2955867 fatcat:ckastwefh5chhamsravandtnx4

Unified Simulation, Perception, and Generation of Human Behavior [article]

Ye Yuan
2022 arXiv   pre-print
We also discuss the lessons learned and our vision for what is next for human behavior modeling.  ...  In this thesis, we take a holistic approach to human behavior modeling and tackle its three essential aspects -- simulation, perception, and generation.  ...  For 3D Locomotion, Transform2Act creates a spider-like agent with long legs.  ... 
arXiv:2204.13678v1 fatcat:ls2kmgifbjbaffgwodgdiilstm

Table of Contents

2022 IEEE Robotics and Automation Letters  
Lueth Dynamics-Aware Metric Embedding: Metric Learning in a Latent Space for Visual Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ...  Rimon Configuration Space Decomposition for Scalable Proxy Collision Checking in Robot Planning and Control . . . . . . . . . . . .  ... 
doi:10.1109/lra.2022.3165102 fatcat:enjzebowe5hn7hsfwklc7nieuy
« Previous Showing results 1 — 15 out of 133 results