Learning latent actions to control assistive robots

Losey, Dylan P.; Jeon, Hong Jun; Li, Mengxi; Srinivasan, K.; Mandlekar, Ajay; Garg, Animesh; Bohg, Jeannette; Sadigh, Dorsa

doi:10.1007/s10514-021-10005-w

Cited by 21 publications

(7 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Notable methods utilize centralized training, which accounts for the actions of other agents through a centralized critic (Lowe et al, 2017;Iqbal & Sha, 2018). Other related approaches add explicit communication channels between agents so that agents can share their policy parameters or gradient updates (Foerster et al, 2016;Losey et al, 2019). These methods are typically concerned with the problem of centralized training for a collection of autonomous agents.…”

Section: Related Workmentioning

confidence: 99%

Learning to Influence Human Behavior with Offline Reinforcement Learning

Hong¹,

Dragan²,

Levine³

2023

Preprint

View full text Add to dashboard Cite

In the real world, some of the most complex settings for learned agents involve interaction with humans, who often exhibit suboptimal, unpredictable behavior due to sophisticated biases. Agents that interact with people in such settings end up in uencing the actions that these people take. Our goal in this work is to enable agents to leverage that in uence to improve the human's performance in collaborative tasks, as the task unfolds. Unlike prior work, we do not assume online training with people (which tends to be too expensive and unsafe), nor access to a high delity simulator of the environment. Our idea is that by taking a variety of previously observed human-human interaction data and labeling it with the task reward, o line reinforcement learning (RL) can learn to combine components of behavior, and uncover actions that lead to more desirable human actions. First, we show that o line RL can learn strategies to in uence and improve human behavior, despite those strategies not appearing in the dataset, by utilizing components of diverse, suboptimal interactions. In addition, we demonstrate that o line RL can learn in uence that adapts with humans, thus achieving long-term coordination with them even when their behavior changes. We evaluate our proposed method with real people in the Overcooked collaborative benchmark domain, and demonstrate successful improvement in human performance.

show abstract

Section: Related Workmentioning

confidence: 99%

Learning to Influence Human Behavior with Offline Reinforcement Learning

Hong¹,

Dragan²,

Levine³

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…And it ensures the correct operation of various parameters in the trajectory autonomous sensing method of multi-degree-of-freedom industrial robot arm in C++ language. 28 Under the above background, the kinova jaco2 7-DOF industrial robot arm is selected as the experimental object. The simulation experimental platform is shown in Figure 7.…”

Section: Simulation Experiments Analysismentioning

confidence: 99%

Autonomous perception method of multi‐degree‐of‐freedom industrial robot arm trajectory

Qian

2023

Adv Control Appl

View full text Add to dashboard Cite

In this study, a novel autonomous sensing method of multi‐degree‐of‐freedom industrial robot arm trajectory is proposed. The research takes the distance sensor to collect environmental data, and takes the point cloud data scanned by 3D laser as the basis. The environment model of multi‐degree‐of‐freedom industrial robotic arm is established by Iterative Closest Point (ICP). Then the target object is calibrated by binocular imaging technology. Subsequently, angle of each joint of multi‐degree‐of‐freedom industrial robotic arm is calculated to determine the spatial attitude of the robotic arm. In addition, 3D LiDAR is installed at the end of the robotic arm, and the end trajectory of multi‐degree‐of‐freedom industrial robotic arm is sensed autonomously by using the optimal function. The proposed method has advantages of high accuracy and short sensing time in autonomous sensing of multi‐degree‐of‐freedom industrial robot arm trajectory.

show abstract

“…Researchers are working on different approaches using skin surface electromyogram–based signals [ 17 ], nonlinear sliding mode control [ 18 ], geometric solution [ 19 ], and variable transformation for flatness geometric property [ 20 ] using collaborative robots to design robots for rehabilitation. Researchers have also followed learning latent actions from task demonstrations [ 21 ], reinforcement learning [ 22 ], digital image processing [ 23 ], and eye tracking–based assistive robot control [ 24 ] approaches for collaborative robots, focusing on assistive applications.…”

Section: Introductionmentioning

confidence: 99%

A Novel Framework for Mixed Reality–Based Control of Collaborative Robot: Development Study

Shahria¹,

Sunny²,

Zarif³

et al. 2022

JMIR Biomed Eng

View full text Add to dashboard Cite

Background Applications of robotics in daily life are becoming essential by creating new possibilities in different fields, especially in the collaborative environment. The potentials of collaborative robots are tremendous as they can work in the same workspace as humans. A framework employing a top-notch technology for collaborative robots will surely be worthwhile for further research. Objective This study aims to present the development of a novel framework for the collaborative robot using mixed reality. Methods The framework uses Unity and Unity Hub as a cross-platform gaming engine and project management tool to design the mixed reality interface and digital twin. It also uses the Windows Mixed Reality platform to show digital materials on holographic display and the Azure mixed reality services to capture and expose digital information. Eventually, it uses a holographic device (HoloLens 2) to execute the mixed reality–based collaborative system. Results A thorough experiment was conducted to validate the novel framework for mixed reality–based control of a collaborative robot. This framework was successfully applied to implement a collaborative system using a 5–degree of freedom robot (xArm-5) in a mixed reality environment. The framework was stable and worked smoothly throughout the collaborative session. Due to the distributed nature of cloud applications, there is a negligible latency between giving a command and the execution of the physical collaborative robot. Conclusions Opportunities for collaborative robots in telerehabilitation and teleoperation are vital as in any other field. The proposed framework was successfully applied in a collaborative session, and it can also be applied in other similar potential applications for robust and more promising performance.

show abstract

Learning latent actions to control assistive robots

Cited by 21 publications

References 33 publications

Learning to Influence Human Behavior with Offline Reinforcement Learning

Learning to Influence Human Behavior with Offline Reinforcement Learning

Autonomous perception method of multi‐degree‐of‐freedom industrial robot arm trajectory

A Novel Framework for Mixed Reality–Based Control of Collaborative Robot: Development Study

Contact Info

Product

Resources

About