An official website of the United States government
Human demonstrations are important in a range of robotics applications, and are created with a variety of input methods. However, the design space for these input methods has not been extensively studied. In this paper, focusing on demonstrations of hand-scale object manipulation tasks to robot arms with two-finger grippers, we identify distinct usage paradigms in robotics that utilize human-to-robot demonstrations, extract abstract features that form a design space for input methods, and characterize existing input methods as well as a novel input method that we introduce, the instrumented tongs. We detail the design specifications for our method and present a user study that compares it against three common input methods: free-hand manipulation, kinesthetic guidance, and teleoperation. Study results show that instrumented tongs provide high quality demonstrations and a positive experience for the demonstrator while offering good correspondence to the target robot.
more »
« less
HERD: Continuous Human-to-Robot Evolution for Learning from Human Demonstration
The ability to learn from human demonstration endows robots with the ability to automate various tasks. However, directly learning from human demonstration is challenging since the structure of the human hand can be very different from the desired robot gripper. In this work, we show that manipulation skills can be transferred from a human to a robot through the use of micro-evolutionary reinforcement learning, where a five-finger human dexterous hand robot gradually evolves into a commercial two-finger-gripper robot, while repeated interacting in a physics simulator to continuously update the policy that is first learned from human demonstration. To deal with the high dimensions of robot parameters, we propose an algorithm for multi-dimensional evolution path searching that allows joint optimization of both the robot evolution path and the policy. Through experiments on human object manipulation datasets, we show that our framework can efficiently transfer the expert human agent policy trained from human demonstrations in diverse modalities to a target commercial robot.
more »
« less
- Award ID(s):
- 2024594
- PAR ID:
- 10366296
- Date Published:
- Journal Name:
- Conference on Robot Learning
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Learning a robot motor skill from scratch is impractically slow; so much so that in practice, learning must typically be bootstrapped using human demonstration. However, relying on human demonstration necessarily degrades the autonomy of robots that must learn a wide variety of skills over their operational lifetimes. We propose using kinematic motion planning as a completely autonomous, sample efficient way to bootstrap motor skill learning for object manipulation. We demonstrate the use of motion planners to bootstrap motor skills in two complex object manipulation scenarios with different policy representations: opening a drawer with a dynamic movement primitive representation, and closing a microwave door with a deep neural network policy. We also show how our method can bootstrap a motor skill for the challenging dynamic task of learning to hit a ball off a tee, where a kinematic plan based on treating the scene as static is insufficient to solve the task, but sufficient to bootstrap a more dynamic policy. In all three cases, our method is competitive with human-demonstrated initialization, and significantly outperforms starting with a random policy. This approach enables robots to to efficiently and autonomously learn motor policies for dynamic tasks without human demonstration.more » « less
-
Jara, C; Borras_Sol, J (Ed.)Deep Reinforcement Learning (DRL) has shown its capability to solve the high degrees of freedom in control and the complex interaction with the object in the multi-finger dexterous in-hand manipulation tasks. Current DRL approaches lack behavior constraints during the learning process, leading to aggressive and unstable policies that are insufficient for safety-critical in-hand manipulation tasks. The centralized learning strategy also limits the flexibility to fine-tune each robot finger's behavior. This work proposes the Finger-specific Multi-agent Shadow Critic Consensus (FMSC) method, which models the in-hand manipulation as a multi-agent collaboration task where each finger is an individual agent and trains the policies for the fingers to achieve a consensus across the critic networks through the Information Sharing (IS) across the neighboring agents and finger-specific stable manipulation objectives based on the state-action occupancy measure, a general utility of DRL that is approximated during the learning process. The methods are evaluated in two in-hand manipulation tasks on the Shadow Hand. The results show that FMSC+IS converges faster in training, achieving a comparable success rate and much better manipulation stability than conventional DRL methods.more » « less
-
Rotation manipulation tasks are a fundamental component of manipulation, however few benchmarks directly measure the limits of a hand's ability to rotate objects. This paper presents two benchmarks for quantitatively measuring the rotation manipulation capabilities of two-fingered hands. These benchmarks exists to augment the Asterisk Test to consider rotation manipulation ability. We propose two benchmarks: the first assesses a hand's limits to rotate objects clockwise and counterclockwise with minimal translation, and the second assesses how rotation manipulation impacts a hand's in-hand translation performance. We demonstrate the utility of these rotation benchmarks using three generic robot hand designs: 1) an asymmetrical two-linked versus one-linked gripper (2v1), 2) a symmetrical two-linked gripper (2v2), and 3) a symmetrical three-linked gripper (3v3). We conclude with a brief comparison between the hand designs and a observations about contact point selection for manipulation tasks, informed from our benchmark results.more » « less
-
Grasp is an integral part of manipulation actions in activities of daily living and programming by demonstration is a powerful paradigm for teaching the assistive robots how to perform a grasp. Since finger configuration and finger force are the fundamental features that need to be controlled during a grasp, using these variables is a natural choice for learning by demonstration. An important question then becomes whether the existing grasp taxonomies are appropriate when one considers these modalities. The goal of our paper is to answer this question by investigating grasp patterns that can be inferred from a static analysis of the grasp data, as the object is securely grasped. Human grasp data is measured using a newly developed data glove. The data includes pressure sensor measurements from eighteen areas of the hand, and measurements from bend sensors placed at finger joints. The pressure sensor measurements are calibrated and mapped into force by employing a novel data-driven approach. Unsupervised learning is used to identify patterns for different grasp types. Multiple clustering algorithms are used to partition the data. When the results are taken in aggregate, 25 human grasp types are reduced to 9 different clusters.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
