Author name cluster

Dmitry Popov

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

2 papers

1 author row

EAAI Journal 2023 Journal Article

Adaptive technique for physical human–robot interaction handling using proprioceptive sensors

Dmitry Popov
Anatol Pashkevich
Alexandr Klimchik

The work focuses on the development of an adaptive technique for the physical interaction handling between a human and a robot, as well as its experimental validation. The proposed technique is based on the deep residual neural network and dedicated finite state machine, where the states are the robot behavior modes and transitions are the switchings between the states that depend on the interaction parameters and characteristics. It ensures the human operator safety and improves the human–robot collaboration performance by implementing various scenarios. In the scope of this technique, the parameters of human–robot interaction are used to select an appropriate robot reaction strategy using data from internal robot sensors only, i. e. proprioceptive sensors. These parameters define the interaction force vector and its application point on the robot surface, which allow to classify the interaction within the set of predefined categories. This classification distinguishes interactions applied at the tool or intermediate link (Tool/Link), having soft or hard nature (Soft/Hard), as well as having different intention (Intl/Accd) or duration (Short/Long). Based on identified category and the current robot state, the algorithm chooses an appropriate robot reaction. To confirm the efficiency the developed technique, an experimental study was conducted, which involved the collaboration between the real industrial manipulator KUKA LBR iiwa and the human operator.

Details DOI

NeurIPS Conference 2021 Conference Paper

Distributed Deep Learning In Open Collaborations

Michael Diskin
Alexey Bukhtiyarov
Max Ryabinin
Lucile Saulnier
Quentin Lhoest
Anton Sinitsin
Dmitry Popov
Dmitry V. Pyrkin

Modern deep learning applications require increasingly more compute to train state-of-the-art models. To address this demand, large corporations and institutions use dedicated High-Performance Computing clusters, whose construction and maintenance are both environmentally costly and well beyond the budget of most organizations. As a result, some research directions become the exclusive domain of a few large industrial and even fewer academic actors. To alleviate this disparity, smaller groups may pool their computational resources and run collaborative experiments that benefit all participants. This paradigm, known as grid- or volunteer computing, has seen successful applications in numerous scientific areas. However, using this approach for machine learning is difficult due to high latency, asymmetric bandwidth, and several challenges unique to volunteer computing. In this work, we carefully analyze these constraints and propose a novel algorithmic framework designed specifically for collaborative training. We demonstrate the effectiveness of our approach for SwAV and ALBERT pretraining in realistic conditions and achieve performance comparable to traditional setups at a fraction of the cost. Finally, we provide a detailed report of successful collaborative language model pretraining with nearly 50 participants.

PDF Details