-
Real-World Modeling of Computation Offloading for Neural Networks with Early Exits and Splits
Authors:
Jan Danek,
Zdenek Becvar,
Adam Janes
Abstract:
We focus on computation offloading of applications based on convolutional neural network (CNN) from moving devices, such as mobile robots or autonomous vehicles, to MultiAccess Edge Computing (MEC) servers via a mobile network. In order to reduce overall CNN inference time, we design and implement CNN with early exits and splits, allowing a flexible partial or full offloading of CNN inference. Thr…
▽ More
We focus on computation offloading of applications based on convolutional neural network (CNN) from moving devices, such as mobile robots or autonomous vehicles, to MultiAccess Edge Computing (MEC) servers via a mobile network. In order to reduce overall CNN inference time, we design and implement CNN with early exits and splits, allowing a flexible partial or full offloading of CNN inference. Through real-world experiments, we analyze an impact of the CNN inference offloading on the total CNN processing delay, energy consumption, and classification accuracy in a practical road sign recognition task. The results confirm that offloading of CNN with early exits and splits can significantly reduce both total processing delay and energy consumption compared to full local processing while not impairing classification accuracy. Based on the results of real-world experiments, we derive practical models for energy consumption and total processing delay related to offloading of CNN with early exits and splits.
△ Less
Submitted 28 May, 2025;
originally announced May 2025.
-
Multi-hop Relaying with Mixed Half and Full Duplex Relays for Offloading to MEC
Authors:
Pavel Mach,
Zdenek Becvar,
Mohammadsaleh Nikooroo
Abstract:
In this paper, we focus on offloading a computing task from a user equipment (UE) to a multi-access edge computing (MEC) server via multi-hop relaying. We assume a general relaying case where relays are energy-constrained devices, such as other UEs, internet of things (IoT) devices, or unmanned aerial vehicles. To this end, we formulate the problem as a minimization of the sum energy consumed by t…
▽ More
In this paper, we focus on offloading a computing task from a user equipment (UE) to a multi-access edge computing (MEC) server via multi-hop relaying. We assume a general relaying case where relays are energy-constrained devices, such as other UEs, internet of things (IoT) devices, or unmanned aerial vehicles. To this end, we formulate the problem as a minimization of the sum energy consumed by the energy-constrained devices under the constraint on the maximum requested time of the task processing. Then, we propose a multi-hop relaying combining half and full duplexes at each individual relay involved in the offloading. We proof that the proposed multi-hop relaying is convex, thus it can be optimized by conventional convex optimization methods. We show our proposal outperforms existing multi-hop relaying schemes in terms of probability that tasks are processed within required time by up to 38\% and, at the same time, decreases energy consumption by up to 28%.
△ Less
Submitted 12 January, 2024;
originally announced January 2024.
-
6G in the Sky: On-Demand Intelligence at the Edge of 3D Networks
Authors:
Emilio Calvanese Strinati,
Sergio Barbarossa,
Taesang Choi,
Antonio Pietrabissa,
Alessandro Giuseppi,
Emanuele De Santis,
Josep Vidal,
Zdenek Becvar,
Thomas Haustein,
Nicolas Cassiau,
Francesca Costanzo,
Junhyeong Kim,
Ilgyu Kim
Abstract:
6G will exploit satellite, aerial and terrestrial platforms jointly to improve radio access capability and to unlock the support of on-demand edge cloud services in the three dimensional space (3D) by incorporating Mobile Edge Computing (MEC) functionalities on aerial platforms and low orbit satellites. This will extend the MEC support to devices and network elements in the sky and will forge a sp…
▽ More
6G will exploit satellite, aerial and terrestrial platforms jointly to improve radio access capability and to unlock the support of on-demand edge cloud services in the three dimensional space (3D) by incorporating Mobile Edge Computing (MEC) functionalities on aerial platforms and low orbit satellites. This will extend the MEC support to devices and network elements in the sky and will forge a space borne MEC enabling intelligent personalized and distributed on demand services. 3D end users will experience the impression of being surrounded by a distributed computer fulfilling their requests in apparently zero latency. In this paper, we consider an architecture providing communication, computation, and caching (C3) services on demand, anytime and everywhere in 3D space, building on the integration of conventional ground (terrestrial) base stations and flying (non-terrestrial) nodes. Given the complexity of the overall network, the C3 resources and the management of the aerial devices need to be jointly orchestrated via AI-based algorithms, exploiting virtualized networks functions dynamically deployed in a distributed manner across terrestrial and non-terrestrial nodes.
△ Less
Submitted 19 October, 2020;
originally announced October 2020.
-
Predicting Device-to-Device Channels from Cellular Channel Measurements: A Learning Approach
Authors:
Mehyar Najla,
Zdenek Becvar,
Pavel Mach,
David Gesbert
Abstract:
Device-to-device (D2D) communication, which enables a direct connection between users while bypassing the cellular channels to base stations (BSs), is a promising way to offload the traffic from conventional cellular networks. In D2D communication, one recurring problem is that, in order to optimally allocate resources across D2D and cellular users, the knowledge of D2D channel gains is needed. Ho…
▽ More
Device-to-device (D2D) communication, which enables a direct connection between users while bypassing the cellular channels to base stations (BSs), is a promising way to offload the traffic from conventional cellular networks. In D2D communication, one recurring problem is that, in order to optimally allocate resources across D2D and cellular users, the knowledge of D2D channel gains is needed. However, such knowledge is hard to obtain at reasonable signaling costs. In this paper, we show this problem can be circumvented by tapping into the information provided by the estimation of the cellular channels between the users and surrounding BSs as this estimation is done anyway for a normal operation of the network. While the cellular and D2D channel gains exhibit independent fast fading behavior, we show that average gains of the cellular and D2D channels share a non-explicit correlation structure, which is rooted into the network topology, terrain, and buildings setup. We propose a machine (deep) learning approach capable of predicting the D2D channel gains from seemingly independent cellular channels. Our results show a high degree of convergence between true and predicted D2D channel gains. The predicted gains allow to reach a near-optimal communication capacity in many radio resource management algorithms.
△ Less
Submitted 17 November, 2019;
originally announced November 2019.
-
Energy Efficiency of Opportunistic Device-to-Device Relaying Under Lognormal Shadowing
Authors:
S. M. Zafaruddin,
Jan Plachy,
Zdenek Becvar,
Amir Leshem
Abstract:
Energy consumption is a major limitation of low power and mobile devices. Efficient transmission protocols are required to minimize an energy consumption of the mobile devices for ubiquitous connectivity in the next generation wireless networks. Opportunistic schemes select a single relay using the criteria of the best channel and achieve a near-optimal diversity performance in a cooperative wirel…
▽ More
Energy consumption is a major limitation of low power and mobile devices. Efficient transmission protocols are required to minimize an energy consumption of the mobile devices for ubiquitous connectivity in the next generation wireless networks. Opportunistic schemes select a single relay using the criteria of the best channel and achieve a near-optimal diversity performance in a cooperative wireless system. In this paper, we study the energy efficiency of the opportunistic schemes for device-to-device communication. In the opportunistic approach, an energy consumed by devices is minimized by selecting a single neighboring device as a relay using the criteria of minimum consumed energy in each transmission in the uplink of a wireless network. We derive analytical bounds and scaling laws on the expected energy consumption when the devices experience log-normal shadowing with respect to a base station considering both the transmission as well as circuit energy consumptions. We show that the protocol improves the energy efficiency of the network comparing to the direct transmission even if only a few devices are considered for relaying. We also demonstrate the effectiveness of the protocol by means of simulations in realistic scenarios of the wireless network.
△ Less
Submitted 2 March, 2018;
originally announced March 2018.
-
Combined shared/dedicated resource allocation for Device-to-Device Communication
Authors:
Pavel Mach,
Zdenek Becvar
Abstract:
Device-to-device (D2D) communication is an effective technology enhancing spectral efficiency and network throughput of contemporary cellular networks. Typically, the users exploiting D2D reuse the same radio resources as common cellular users (CUEs) that communicate through a base station. This mode is known as shared mode. Another option is to dedicate specific amount of resources exclusively fo…
▽ More
Device-to-device (D2D) communication is an effective technology enhancing spectral efficiency and network throughput of contemporary cellular networks. Typically, the users exploiting D2D reuse the same radio resources as common cellular users (CUEs) that communicate through a base station. This mode is known as shared mode. Another option is to dedicate specific amount of resources exclusively for the D2D users in so-called a dedicated mode. In this paper, we propose novel combined share/dedicated resource allocation scheme enabling the D2D users to utilize the radio resources in both modes simultaneously. To that end, we propose a graph theory-based framework for efficient resource allocation. Within this framework, neighborhood relations between the cellular users and the D2D users and between the individual D2D users are derived to form graphs. Then, the graphs are decomposed into subgraphs to identify resources, which can be reused by other users so that capacity of the D2D users is maximized. The results show that the sum D2D capacity is increased from 1.67 and 2.5 times (depending on a density of D2D users) when compared to schemes selecting only between shared or dedicated modes.
△ Less
Submitted 6 November, 2017;
originally announced November 2017.
-
Combination of Visible Light and Radio Frequency Bands for Device-to-Device Communication
Authors:
Pavel Mach,
Zdenek Becvar,
Mehyar Najla,
Stanislav Zvanovec
Abstract:
Future mobile networks are supposed to serve high data rates required by users. To accommodate the high data rates, a direct communication between nearby mobile terminals (MTs), known as Device-to-device (D2D) communication, can be exploited. Furthermore, a communication in high frequency bands, such as, visible light communication (VLC), is also foreseen as an enabler for the high data rates. In…
▽ More
Future mobile networks are supposed to serve high data rates required by users. To accommodate the high data rates, a direct communication between nearby mobile terminals (MTs), known as Device-to-device (D2D) communication, can be exploited. Furthermore, a communication in high frequency bands, such as, visible light communication (VLC), is also foreseen as an enabler for the high data rates. In conventional D2D communication, pairs of the MTs should reuse the same frequencies to keep a high spectral efficiency of the system. However, this implies either interference among the D2D pairs or utilization of complex and advanced resource allocation algorithms. In this paper, we propose a new concept for D2D communication combining VLC and RF technologies in order to maximize capacity of the system. This paper provides an analysis of the operational limits for the proposed concept and investigates capacity gains introduced by the combined usage of RF and VLC bands for D2D. Moreover, we discuss several practical issues related to the proposed RF-VLC D2D concept. Performed analyses show that the RF--VLC D2D is able to improve the capacity in an indoor scenario by a factor of 4.1 and 1.5 when compared to stand-alone RF D2D and VLC D2D, respectively.
△ Less
Submitted 14 June, 2017;
originally announced September 2017.
-
Mobile Edge Computing: A Survey on Architecture and Computation Offloading
Authors:
Pavel Mach,
Zdenek Becvar
Abstract:
Technological evolution of mobile user equipments (UEs), such as smartphones or laptops, goes hand-in-hand with evolution of new mobile applications. However, running computationally demanding applications at the UEs is constrained by limited battery capacity and energy consumption of the UEs. Suitable solution extending the battery life-time of the UEs is to offload the applications demanding hug…
▽ More
Technological evolution of mobile user equipments (UEs), such as smartphones or laptops, goes hand-in-hand with evolution of new mobile applications. However, running computationally demanding applications at the UEs is constrained by limited battery capacity and energy consumption of the UEs. Suitable solution extending the battery life-time of the UEs is to offload the applications demanding huge processing to a conventional centralized cloud (CC). Nevertheless, this option introduces significant execution delay consisting in delivery of the offloaded applications to the cloud and back plus time of the computation at the cloud. Such delay is inconvenient and make the offloading unsuitable for real-time applications. To cope with the delay problem, a new emerging concept, known as mobile edge computing (MEC), has been introduced. The MEC brings computation and storage resources to the edge of mobile network enabling to run the highly demanding applications at the UE while meeting strict delay requirements. The MEC computing resources can be exploited also by operators and third parties for specific purposes. In this paper, we first describe major use cases and reference scenarios where the MEC is applicable. After that we survey existing concepts integrating MEC functionalities to the mobile networks and discuss current advancement in standardization of the MEC. The core of this survey is, then, focused on user-oriented use case in the MEC, i.e., computation offloading. In this regard, we divide the research on computation offloading to three key areas: i) decision on computation offloading, ii) allocation of computing resource within the MEC, and iii) mobility management. Finally, we highlight lessons learned in area of the MEC and we discuss open research challenges yet to be addressed in order to fully enjoy potentials offered by the MEC.
△ Less
Submitted 13 March, 2017; v1 submitted 17 February, 2017;
originally announced February 2017.
