HASP: Hierarchical Asynchronous Parallelism for Multi-NN Tasks DOI
Hongyi Li, Songchen Ma, Taoyi Wang

et al.

IEEE Transactions on Computers, Journal Year: 2023, Volume and Issue: 73(2), P. 366 - 379

Published: Nov. 7, 2023

The rapid development of deep learning has propelled many real-world artificial intelligence applications. Many these applications integrate multiple neural networks (multi-NN) to cater various functionalities. There are two challenges multi-NN acceleration: (1) competition for shared resources becomes a bottleneck, and (2) heterogeneous workloads exhibit remarkably different computing-memory characteristics synchronization requirements. Therefore, resource isolation fine-grained allocation each task fundamental requirements computing systems. Although number acceleration technologies have been explored, few can completely fulfill both requirements, especially mobile scenarios. This paper reports Hierarchical Asynchronous Parallel Model (HASP) enhance performance meet HASP be implemented on multicore processor that adopts Multiple Instruction Data (MIMD) or Single Thread (SIMT) architectures, with minor adaptive modification needed. Further, prototype chip is developed validate the hardware effectiveness this design. A corresponding mapping strategy also developed, allowing proposed architecture simultaneously promote utilization throughput. With same workload, demonstrates 3.62 $\boldsymbol{\times}$ , 3.51 higher throughput over Planaria 8.68 2.61 Jetson AGX Orin MobileNet-V1 ResNet50, respectively.

Language: Английский

A multimodal teleoperation interface for human-robot collaboration DOI
Weiyong Si,

Tianjian Zhong,

Ning Wang

et al.

Published: March 15, 2023

Human-robot collaboration provides an effective approach to combine human intelligence and the autonomy of robots, which can improve safety efficiency robot. However, developing intuitive immersive human-robot interface with multimodal feedback for interaction is still challenging. In this paper, we developed a multimodal-based involve humans in loop. The Unity-based virtual reality (VR) environment, including robot manipulator its working was simulate real environment robots. We integrated digital twin mechanism VR development, corresponding model physical task. could visualize visual haptic through sensors robot, friendly teleoperating operators. conduct user study experiments based on NASA Task Load Index, contact scanning result shows that proposed improved by 31.8% terms cognitive workload, comparing commercial teleportation device Touch X.

Language: Английский

Citations

6
A Novel Interval Type-2 Fuzzy Classifier Based on Explainable Neural Network for Surface Electromyogram Gesture Recognition DOI
Shuai Lv, Zhijun Li, Jin Huang

et al.

IEEE Transactions on Human-Machine Systems, Journal Year: 2023, Volume and Issue: 53(6), P. 955 - 964

Published: Sept. 12, 2023

The existing hand gesture classification research based on surface electromyogram (sEMG) faces the challenges of low accuracy, weak real-time ability, robustness, few categories, and lack explainability. In this article, we investigate how to classify sEMG signals for grasp recognition human–robot interaction consider these issues. A novel interval type-2 (IT2) fuzzy classifier explainable neural network is proposed recognition. Based fully connected network, adaptive moment estimation applied tune antecedent parameters. Ninapro data adopted test performance model, which realizes 52 gestures achieves 95.04% categorization accuracy. Moreover, grasping experiments are conducted computer, communication, consumer electronics (3C) experiment platform ability in real scenarios. recognizes six gestures. results 3C show that method 99.4% offline training accuracy as well 96.07% online Meanwhile, 89.4% can be obtained within 0.5 s. overall demonstrate great potential real-world applications, such human intent detection manipulator control.

Language: Английский

Citations

6
Reshaping Wearable Robots Using Fuzzy Intelligence: Integrating Type-2 Fuzzy Decision, Intelligent Control, and Origami Structure DOI
Shiyuan Bian, Jonathan M. Garibaldi, Zhijun Li

et al.

IEEE Transactions on Fuzzy Systems, Journal Year: 2023, Volume and Issue: 31(11), P. 3741 - 3761

Published: May 15, 2023

Currently, type-2 fuzzy systems, control strategies, and origami structures have been used in robots to assist the comfortable smooth operation of such robots. Recent advances these various fields employed improve key technologies wearable robots, as providing more efficient decision making, improved maneuverability, increased intelligence, lightweight structures. The current highlighted potential for methods, both separately combination with each other, achieve further significant advances. Hence, this article summarizes latest research results three aspects, elaborates on some challenges that remain, discusses development directions future.

Language: Английский

Citations

5
Modeling and Control of an Octopus Inspired Soft Arm under Prescribed Spatial Motion Constraints DOI

Jie Ma,

Zhiji Han, Zhijie Liu

et al.

Journal of Intelligent & Robotic Systems, Journal Year: 2023, Volume and Issue: 109(4)

Published: Dec. 1, 2023

Language: Английский

Citations

5
HASP: Hierarchical Asynchronous Parallelism for Multi-NN Tasks DOI
Hongyi Li, Songchen Ma, Taoyi Wang

et al.

IEEE Transactions on Computers, Journal Year: 2023, Volume and Issue: 73(2), P. 366 - 379

Published: Nov. 7, 2023

The rapid development of deep learning has propelled many real-world artificial intelligence applications. Many these applications integrate multiple neural networks (multi-NN) to cater various functionalities. There are two challenges multi-NN acceleration: (1) competition for shared resources becomes a bottleneck, and (2) heterogeneous workloads exhibit remarkably different computing-memory characteristics synchronization requirements. Therefore, resource isolation fine-grained allocation each task fundamental requirements computing systems. Although number acceleration technologies have been explored, few can completely fulfill both requirements, especially mobile scenarios. This paper reports Hierarchical Asynchronous Parallel Model (HASP) enhance performance meet HASP be implemented on multicore processor that adopts Multiple Instruction Data (MIMD) or Single Thread (SIMT) architectures, with minor adaptive modification needed. Further, prototype chip is developed validate the hardware effectiveness this design. A corresponding mapping strategy also developed, allowing proposed architecture simultaneously promote utilization throughput. With same workload, demonstrates 3.62 $\boldsymbol{\times}$ , 3.51 higher throughput over Planaria 8.68 2.61 Jetson AGX Orin MobileNet-V1 ResNet50, respectively.

Language: Английский

Citations

5