Network Offloading Policies for Cloud Robotics: A Learning-Based Approach

Chinchali, Sandeep; Sharma, Anita; Harrison, J. Michael; Elhafsi, Amine; Kang, Daniel; Pergament, Evgenya; Cidon, Eyal; Pavone, Marco

doi:10.15607/rss.2019.xv.063

Cited by 53 publications

(27 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, a lightweight drone fleet may not have the power or weight budget to conduct massive computations on board the apparatus, but with a wide enough channel bandwidth and sufficiently fast data rate, real time computations for extremely complex tasks, such as contextual awareness, vision, and perception may be carried out at a fixed base station or edge server that is in wireless connection and supporting real time cognition for the drone fleet. Robots, autonomous vehicles, and other machines may be similarly designed to exploit cognitive processing performed remotely from the machine using wireless, with the ability to perform tasks without the benefit of local cognition on the platform [45], [46].…”

Section: A Wireless Cognitionmentioning

confidence: 99%

Wireless Communications and Applications Above 100 GHz: Opportunities and Challenges for 6G and Beyond

et al. 2019

View full text Add to dashboard Cite

Frequencies from 100 GHz to 3 THz are promising bands for the next generation of wireless communication systems because of the wide swaths of unused and unexplored spectrum. These frequencies also offer the potential for revolutionary applications that will be made possible by new thinking, and advances in devices, circuits, software, signal processing, and systems. This paper describes many of the technical challenges and opportunities for wireless communication and sensing applications above 100 GHz, and presents a number of promising discoveries, novel approaches, and recent results that will aid in the development and implementation of the sixth generation (6G) of wireless networks, and beyond. This paper shows recent regulatory and standard body rulings that are anticipating wireless products and services above 100 GHz and illustrates the viability of wireless cognition, hyper-accurate position location, sensing, and imaging. This paper also presents approaches and results that show how long distance mobile communications will be supported to above 800 GHz since the antenna gains are able to overcome airinduced attenuation, and present methods that reduce the computational complexity and simplify the signal processing used in adaptive antenna arrays, by exploiting the Special Theory of Relativity to create a cone of silence in over-sampled antenna arrays that improve performance for digital phased array antennas. Also, new results that give insights into power efficient beam steering algorithms, and new propagation and partition loss models above 100 GHz are given, and promising imaging, array processing, and position location results are presented. The implementation of spatial consistency at THz frequencies, an important component of channel modeling that considers minute changes and correlations over space, is also discussed. This paper offers the first in-depth look at the vast applications of THz wireless products and applications and provides approaches for how to reduce power and increase performance across several problem domains, giving early evidence that THz techniques are compelling and available for future wireless communications. INDEX TERMS mmWave, millimeter wave, 5G, D-band, 6G, channel sounder, propagation measurements, Terahertz (THz), array processing, imaging, scattering theory, cone of silence, digital phased arrays, digital beamformer, signal processing for THz, position location, channel modeling, THz applications, wireless cognition, network offloading. I. INTRODUCTION The tremendous funding and research efforts invested in millimeter wave (mmWave) wireless communications, and The associate editor coordinating the review of this manuscript and approving it for publication was Thomas Kuerner. the early success of 5G trials and testbeds across the world, ensure that commercial widespread 5G wireless networks will be realized by 2020 [1]. The use of mmWave in 5G wireless communication will solve the spectrum shortage in current 4G cellular communication systems that operate

show abstract

Section: A Wireless Cognitionmentioning

confidence: 99%

Wireless Communications and Applications Above 100 GHz: Opportunities and Challenges for 6G and Beyond

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Notably, the aforementioned algorithms were all designed for mobile devices and mobile-specific applications using MEC, whereas our proposed algorithm is designed from the robotics perspective and validated with a robotic application using cloud computing. Some researchers have applied DRL in application offloading solutions in cloud robotics; for example, Chicachali et al [46] formulated the offloading problem as a sequential decision-making problem and used deep reinforcement learning for object detection applications, and their findings suggest that RL is likely an effective choice for optimizing offloading decision policies. Another prior study proposed a resource allocation scheme based on RL that allowed the cloud to decide whether a request should be accepted and the amount of resources to be allocated to the application; this work also demonstrated better performance of RL algorithms relative to other greedy resource allocation scheme [47].…”

Section: A Related Workmentioning

confidence: 99%

A Deep Reinforcement Learning-Based Dynamic Computational Offloading Method for Cloud Robotics

Penmetcha

Min

2021

IEEE Access

View full text Add to dashboard Cite

Robots come with a variety of computing capabilities, and running computationallyintense applications on robots is sometimes challenging on account of limited onboard computing, storage, and power capabilities. Meanwhile, cloud computing provides on-demand computing capabilities, and thus combining robots with cloud computing can overcome the resource constraints robots face. The key to effectively offloading tasks is an application solution that does not underutilize the robot's own computational capabilities and makes decisions based on crucial cost parameters such as latency and CPU availability. In this paper, we formulate the application offloading problem as a Markovian decision process and propose a deep reinforcement learning-based deep Q-network (DQN) approach. The statespace is formulated with the assumption that input data size directly impacts application execution time. The proposed algorithm is designed as a continuous task problem with discrete action space; i.e., we apply a choice of action at each time step and use the corresponding outcome to train the DQN to acquire the maximum rewards possible. To validate the proposed algorithm, we designed and implemented a robot navigation testbed. The results demonstrated that for the given state-space values, the proposed algorithm learned to take appropriate actions to reduce application latency and also learned a policy that takes actions based on input data size. Finally, we compared the proposed DQN algorithm with a long short-term memory (LSTM) algorithm in terms of accuracy. When trained and validated on the same dataset, the proposed DQN algorithm obtained at least 9 percentage points greater accuracy than the LSTM algorithm.INDEX TERMS Cloud robotics, deep reinforcement learning, deep Q-networks (DQN), AWS, neural networks, application offloading, robot navigation.

show abstract

“…This has motivated several models for appropriate resource allocation and service provisioning [34]. Chinchali et al use a deep reinforcement learning strategy to offload robot sensing tasks over the network [56]. Nan et al present a fog robotic system for dynamic visual servoing with an ayschronous heartbeat signal [57].…”

Section: Flexibility Of Resource Placement and Allocationmentioning

confidence: 99%

A Fog Robotics Approach to Deep Robot Learning: Application to Object Recognition and Grasp Planning in Surface Decluttering

Tanwani

Mor

Kubiatowicz

et al. 2019

2019 International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

The growing demand of industrial, automotive and service robots presents a challenge to the centralized Cloud Robotics model in terms of privacy, security, latency, bandwidth, and reliability. In this paper, we present a 'Fog Robotics' approach to deep robot learning that distributes compute, storage and networking resources between the Cloud and the Edge in a federated manner. Deep models are trained on non-private (public) synthetic images in the Cloud; the models are adapted to the private real images of the environment at the Edge within a trusted network and subsequently, deployed as a service for lowlatency and secure inference/prediction for other robots in the network. We apply this approach to surface decluttering, where a mobile robot picks and sorts objects from a cluttered floor by learning a deep object recognition and a grasp planning model. Experiments suggest that Fog Robotics can improve performance by sim-to-real domain adaptation in comparison to exclusively using Cloud or Edge resources, while reducing the inference cycle time by 4× to successfully declutter 86% of objects over 213 attempts.

show abstract

Network Offloading Policies for Cloud Robotics: A Learning-Based Approach

Cited by 53 publications

References 27 publications

Wireless Communications and Applications Above 100 GHz: Opportunities and Challenges for 6G and Beyond

Wireless Communications and Applications Above 100 GHz: Opportunities and Challenges for 6G and Beyond

A Deep Reinforcement Learning-Based Dynamic Computational Offloading Method for Cloud Robotics

A Fog Robotics Approach to Deep Robot Learning: Application to Object Recognition and Grasp Planning in Surface Decluttering

Contact Info

Product

Resources

About