Filters








694 Hits in 3.4 sec

DART: Noise Injection for Robust Imitation Learning [article]

Michael Laskey, Jonathan Lee, Roy Fox, Anca Dragan, Ken Goldberg
2017 arXiv   pre-print
One approach to Imitation Learning is Behavior Cloning, in which a robot observes a supervisor and infers a control policy.  ...  We propose a new algorithm, DART (Disturbances for Augmenting Robot Trajectories), that collects demonstrations with injected noise, and optimizes the noise level to approximate the error of the robot's  ...  Collaborative Human-Robot Learning (SCHooL) Project, NSF National Robotics Initiative Award 1734633.  ... 
arXiv:1703.09327v2 fatcat:ax27iisqqbdxhmkg22nsmv3qy4

Disturbance-Injected Robust Imitation Learning with Task Achievement [article]

Hirotaka Tahara, Hikaru Sasaki, Hanbit Oh, Brendan Michael, Takamitsu Matsubara
2022 arXiv   pre-print
Robust imitation learning using disturbance injections overcomes issues of limited variation in demonstrations.  ...  To address this issue, this paper proposes a novel imitation learning framework that combines both policy robustification and optimal demonstration learning.  ...  INTRODUCTION Behavior Cloning (BC) [1] is widely used in robotics as an imitation learning (IL) method [2] , [3] to leverage human demonstrations for learning control policies.  ... 
arXiv:2205.04195v1 fatcat:cq2p72pntvhh5hs5p6zhh7iz7y

Recruitment-imitation Mechanism for Evolutionary Reinforcement Learning [article]

Shuai Lü and Shuai Han and Wenbo Zhou and Junwei Zhang
2019 arXiv   pre-print
In this paper, we propose Recruitment-imitation Mechanism (RIM) for evolutionary reinforcement learning, a scalable framework that combines advantages of the three methods mentioned above.  ...  Reinforcement learning, evolutionary algorithms and imitation learning are three principal methods to deal with continuous control tasks.  ...  Performance of components of different algorithms The externally injected individuals accepted by RIM's population are imitation learning individuals, while the externally injected individuals accepted  ... 
arXiv:1912.06310v1 fatcat:zsoydogczjd7ho22aml5weqae4

A Novel FPGA-based Evolvable Hardware System Based on Multiple Processing Arrays

Angel Gallego, Javier Mora, Andres Otero, Ruben Salvador, Eduardo de la Torre, Teresa Riesgo
2013 2013 IEEE International Symposium on Parallel & Distributed Processing, Workshops and Phd Forum  
In this paper, an architecture based on a scalable and flexible set of Evolvable Processing arrays is presented.  ...  The evolvable HW array shown is tailored for window-based image processing applications.  ...  So, a given filter can learn online from another one in the chain, just by imitating it.  ... 
doi:10.1109/ipdpsw.2013.56 dblp:conf/ipps/GallegoMOSTR13 fatcat:nhozwk6b4fbink7eghxypbsd3i

Sample-Efficient Imitation Learning via Generative Adversarial Nets [article]

Lionel Blondé, Alexandros Kalousis
2019 arXiv   pre-print
GAIL is a recent successful imitation learning architecture that exploits the adversarial training procedure introduced in GANs.  ...  We dramatically shrink the amount of interactions with the environment necessary to learn well-behaved imitation policies, by up to several orders of magnitude.  ...  This observation directly echoes noise-injection techniques from the GAN literature.  ... 
arXiv:1809.02064v3 fatcat:xxihg6wl2bdy3kci6kzhoj5lli

Fighting Copycat Agents in Behavioral Cloning from Observation Histories [article]

Chuan Wen, Jierui Lin, Trevor Darrell, Dinesh Jayaraman, Yang Gao
2020 arXiv   pre-print
Imitation learning trains policies to map from input observations to the actions that an expert would choose.  ...  In our experiments, our approach improves performance significantly across a variety of partially observed imitation learning tasks.  ...  Require: learning rate schedule of E and F , l EF ; learning rate schedule of D, l D ; embedding noise std σ; batch size m; number of frames in the imitation learning H for number of training iterations  ... 
arXiv:2010.14876v1 fatcat:4rz7qgmn2vdgxbew5hzhiymygi

Exploring the Limitations of Behavior Cloning for Autonomous Driving [article]

Felipe Codevilla, Eder Santana, Antonio M. López, Adrien Gaidon
2019 arXiv   pre-print
In contrast, imitation learning can, in theory, leverage data from large fleets of human-driven cars.  ...  In this paper, we propose a new benchmark to experimentally investigate the scalability and limitations of behavior cloning.  ...  Noise Distribution During training data collection, 20% of the time we injected noise into expert's steering signal.  ... 
arXiv:1904.08980v1 fatcat:2736atclgfbhrhg72g2tzu3nwy

Few-Shot Bayesian Imitation Learning with Logical Program Policies [article]

Tom Silver, Kelsey R. Allen, Alex K. Lew, Leslie Pack Kaelbling, Josh Tenenbaum
2019 arXiv   pre-print
We argue that the proposed method is an apt choice for tasks that have scarce training data and feature significant, structured variation between task instances.  ...  Humans can learn many novel tasks from a very small number (1--5) of demonstrations, in stark contrast to the data requirements of nearly tabula rasa deep learning methods.  ...  acknowledge support from NSF grants 1523767 and 1723381; from ONR grant N00014-13-1-0333; from AFOSR grant FA9550-17-1-0165; from ONR grant N00014-18-1-2847; from Honda Research; and from the Center for  ... 
arXiv:1904.06317v2 fatcat:o5yo6zyljfdxffdowcxugljita

Reinforcement Learning for Battery Energy Storage Dispatch augmented with Model-based Optimizer [article]

Gayathri Krishnamoorthy, Anamika Dubey
2021 arXiv   pre-print
Specifically, we propose imitation learning based improvements in deep reinforcement learning (DRL) methods to solve the OPF problem for a specific case of battery storage dispatch in the power distribution  ...  The proposed imitation learning algorithm uses the approximate optimal solutions obtained from a linearized model-based OPF solver to provide a good initial policy for the DRL algorithms while improving  ...  Such formalisms aid learning by providing agents examples to imitate.  ... 
arXiv:2109.01659v1 fatcat:b5hydhxtibcpvmtmpsjiaxlhey

Understanding the Spread of COVID-19 Epidemic: A Spatio-Temporal Point Process View [article]

Shuang Li, Lu Wang, Xinyun Chen, Yixiang Fang, Yan Song
2021 arXiv   pre-print
We further adopt a generative adversarial imitation learning framework to learn the model parameters.  ...  In comparison with the traditional likelihood-based learning methods, this imitation learning framework does not need to prespecify an intensity function, which alleviates the model-misspecification.  ...  A random noise vector is injected as part of the inputs to stimulate exploration.  ... 
arXiv:2106.13097v1 fatcat:jkfrjbolxnfnne5yxnelw2eil4

Risk-Sensitive Generative Adversarial Imitation Learning [article]

Jonathan Lacotte and Mohammad Ghavamzadeh and Yinlam Chow and Marco Pavone
2018 arXiv   pre-print
We consider the generative adversarial approach to imitation learning (GAIL) and derive an optimization problem for our formulation, which we call it risk-sensitive GAIL (RS-GAIL).  ...  We first formulate our risk-sensitive imitation learning setting.  ...  In the OpenAI classical control tasks, we inject stochasticity to the system by adding noise to the actions, which in turn adds noise to both the reward function and the transitions.  ... 
arXiv:1808.04468v2 fatcat:ilsz4662xjfitihojjmz2vzw4i

Self-Supervised Correspondence in Visuomotor Policy Learning [article]

Peter Florence, Lucas Manuelli, Russ Tedrake
2019 arXiv   pre-print
In this paper we explore using self-supervised correspondence for improving the generalization performance and sample efficiency of visuomotor policy learning.  ...  amounts of data: using imitation learning, we demonstrate extensive hardware validation on challenging manipulation tasks with as few as 50 demonstrations.  ...  While other works have shown that injecting noise into the dynamics either during imitation learning [33] or sim-to-real transfer [34] can alleviate cascading errors, we provide a simple method based  ... 
arXiv:1909.06933v1 fatcat:53b2u34y4jggzj65fzutqmyxx4

Keyframe-Focused Visual Imitation Learning [article]

Chuan Wen, Jierui Lin, Jianing Qian, Yang Gao, Dinesh Jayaraman
2021 arXiv   pre-print
Imitation learning trains control policies by mimicking pre-recorded expert demonstrations.  ...  Recent solutions ranging from causal graph learning to deep information bottlenecks have shown promising results, but failed to scale to realistic settings such as visual imitation.  ...  Note that CARLA100 is a particularly challenging testbed because it applies the best known techniques for alleviating distributional shift issues in offline imitation, namely, noise injection (Laskey  ... 
arXiv:2106.06452v1 fatcat:z4xwynljb5d5nopdzcqxs7ijci

RRL: Resnet as representation for Reinforcement Learning [article]

Rutav Shah, Vikash Kumar
2021 arXiv   pre-print
The appeal of RRL lies in its simplicity in bringing together progress from the fields of Representation Learning, Imitation Learning, and Reinforcement Learning.  ...  We propose RRL: Resnet as representation for Reinforcement Learning -- a straightforward yet effective approach that can learn complex behaviors directly from proprioceptive inputs.  ...  In this section, we outline related works leveraging representation learning for visual reinforcement and imitation learning.  ... 
arXiv:2107.03380v3 fatcat:bqv2dd3mk5csxcet7ptzz6q7xq

Scalable Synthesis of Verified Controllers in Deep Reinforcement Learning [article]

Zikang Xiong, Suresh Jagannathan
2021 arXiv   pre-print
There has been significant recent interest in devising verification techniques for learning-enabled controllers (LECs) that manage safety-critical systems.  ...  These methods, however, have shown to have significant scalability limitations because verification costs grow as problem dimensionality and objective complexity increase.  ...  Another line of work explores verifiability by applying imitation learning techniques on the subject networks [34, 3, 33] .  ... 
arXiv:2104.10219v2 fatcat:wmghro6mpzcmboj2ai5ihlplju
« Previous Showing results 1 — 15 out of 694 results