A Context‑Aware and Energy‑Efficient Edge Computing Framework

for Low‑Latency Communication in Autonomous Vehicles with

Real‑World Validation and Safety‑Centric Task Prioritization

S. Kannadhasan

, Digant Hemant Raval

, K. Jayalakshmi

, P. Mathiyalagan

M. Soma Sabitha

and Kaviyan S.

Department of Electronics and Communication Engineering, Study World College of Engineering, Coimbatore - 641 105,

Tamil Nadu, India

Department of Mechatronics, G H Patel College of Engineering and Technology, Vallabh Vidyanagar, Gujarat, India

Department of Computer Science and Engineering, R.M.K Engineering College, Kavaraipettai, Tamil Nadu, India

Department of Mechanical Engineering, J.J. College of Engineering and Technology, Tiruchirappalli, Tamil Nadu, India

Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, Telangana, India

Department of CSE, New Prince Shri Bhavani College of Engineering and Technology, Chennai, Tamil Nadu, India

Keywords: Edge Computing, Low‑Latency Communication, Autonomous Vehicles, Context‑Aware Systems,

Safety‑Critical Applications.

Abstract: AVs need ultra-low latency communications to allow for fast decision making for passenger safety in rapidly

changing when the traffic. Legacy cloud processing incurs unsatisfactory latency while most of the current

edge computing approaches suffer from limited flexibility, energy efficiency and practical deployment

experience. This paper presents a context-aware energy-efficient edge computing architecture to realize low-

latency communication for autonomous vehicles. Dynamic task scheduling, federated edge collaboration and

lightweight AI models are jointly used in the framework to make sure that it can realize real-time perception

for safety-critical scenes. The effectiveness of the protocol is verified by actual traffic simulation and multi-

scenario tests, which show reduced response time and improved reliability over different vehicle scenarios.

By combining edge intelligence with safety-aware scheduling and situational awareness, this work addresses

significant deficiencies of the existing generation of vehicle-to-vehicle communication systems and offers an

efficient, security-aware, and low-latency solution for the future intelligent transportation.

1 INTRODUCTION

The fast development of autonomous vehicles (AVs)

have been revolutionizing the intelligent

transportation paradigm, and hence, results in highly

latency-sensitive communication systems that

provide ultra-high-speed data-rate transmission.

Since AVs heavily depend on real-time perception,

computing, and control as well as decision capability

and latency is clearly the critical issue that must be

resolved in order to develop AVs to meet the

necessary road safety and driving performance

requirements. Classical cloud-based architectures, on

the other hand, although computationally intensive,

frequently break down in their application to solve

hard real-time latency demands because of the

intrinsic delay introduced by data transmission over a

long distance. To overcome this challenge, edge

computing has been identified as a promising

paradigm, which locates computation resources near

the vehicle to perform fast data processing and

immediate decision execution.

Though edge computing has promised a bright

future, current solutions suffer from many drawbacks

such as static resource allocation, lack of support for

safety-critical task differentiation and se nse of

context. Furthermore, most works rely on simulation

environments that do not reproduce the unpredictable

and dynamic behavior of the traffic in real situations.

Energy efficiency, scalability across diverse

environments, and secureness strategies are also

insufficiently explored in the existing edge-assisted

AV communication models. Their limitations leave

Kannadhasan, S., Raval, D. H., Jayalakshmi, K., Mathiyalagan, P., Sabitha, M. S. and S., K.

A Contextâ

SAware and Energyâ

SEfﬁcient Edge Computing Framework for Lowâ

SLatency Communication in Autonomous Vehicles with Realâ

SWorld Validation and Safetyâ

SCentric

Task Prioritization.

DOI: 10.5220/0013857000004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 1, pages

27-33

ISBN: 978-989-758-777-1

a great research vacuum which should be filled by a

much more adjustable and systemic method.

In this paper, we propose a new context-aware edge

computing model to provide low-latency

communication for autonomous vehicles. The

proposed system combines real-time sensor data with

environmental context, safeguards safety-critical

maneuvers, and minimizes computational load to

improve both safety and performance. In addition, the

model is verified over practical cellular data and

traffic, which suggests the potential of the proposed

approach. The research seeks to achieve a new

milestone in vehicular communication autonomy by

providing fast and reliable, intelligent decision-

making at the network edge.

1.1 Problem Statement

The widespread deployment of autonomous vehicles

(AVs) in modern transportation systems has become

a critical demand for low-delay ultra-reliable

communication to enable time-sensitive services

including object detection, path planning and

collision avoidance. Autonomous agents require

millisecond-scale processing and reaction times to

plan and execute in real-time in dynamic

environments; it takes little — just milliseconds — to

determine whether the agent is safe, successful, or has

failed. Classical cloud-based architectures, while

strong in terms of computational resources, suffer

delay due to the large delay involved by the

computation being performed off the moving

vehicles, thus they are not suitable for low-latency

vehicular applications.

In order to address this issue, edge computing has

come into picture as a remedy, where data processing

takes place close to the source. Nonetheless, current

edge computing systems for AVs have some

significant shortcomings. Most of these existing

solutions are based on fixed-resource allocation

models, which cannot adjust to changing traffic

conditions and varying environmental settings. In

addition, such systems typically consider all task

equally, and do not have any mechanism for

intelligent prioritization of safety critical operations

like emergency braking or obstacle avoidance. The

lack of contextual awareness results that decisions are

taken without taking into account environmental

factors such as road wetness or dryness, the weather,

or the density of traffic that are necessary to guarantee

security and efficiency in driving.

A further issue is the too strong focus on synthetic

validation simulations, which cannot fully capture the

complexity and randomness of reality. Moreover,

power consumption is often neglected, which is

problematic in terms of deployment in battery-

operated edge nodes or resource-restricted scenarios.

The challenges of security and privacy are also not

well addressed even though vehicular data is

sensitive and edge nodes are susceptible to attacks.

These limitations underscore an urgent need for a

holistic, intelligent, and efficient edge computing

paradigm that can support low-latency

communications, while being flexible to adapt to the

real-world conditions, prioritize safety-critical tasks

in a dynamic manner, cater to energy-constrained

environments, and offer empirical evaluations over

realistic settings. The solutions to these issues are

essential for the reliability, scalability, and

trustworthiness of autonomous vehicle systems in

the next generation intelligent transportation

networks.

2 LITERATURE SURVEY

The requirement for low-latency communication in

AVs has also motivated a significant amount of

research around incorporating edge computing for

processing collected data near the source, resulting in

more informed decisions with faster response times.

Alghamdi and Baz (2021) conducted a foundational

review of edge computing architectures for AVs,

drawing attention to the movement from cloud-based

to decentralized edge systems but recognized that

practical applications are still in their infancy. Wang

et al. (2022) presented a task offloading in

consideration of delay problem for vehicular edge

computing, which can achieve lower data processing

latency in static scene, but suffer from lack

adaptation to the non-deterministic real-world

scenario.

Sun et al. (2021) developed a deep reinforcement

learning-based framework for edge decision-making;

however, the large training cost and complexity of the

model make it unsuitable for real-time vehicular

systems. Similarly, Shen et al. (2023) presented a

cooperative edge intelligence architecture for urban

connected vehicles, by delivering a flexible

architecture viewpoint, although their model was

theoretical having no deployment statistics. In the

context of hybrid computing, Ahmed and Kim (2021)

introduced edge-cloud integrations for AVs that

demonstrate superiority in terms of latency but do not

quantify its energy cost or load scalability.

Rahman and Mehedi (2022) proposed a bespoke

vehicular edge computing framework for V2X

communication to speed up vehicle safety-critical

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

applications, but their approach was only evaluated in

an urban scenario and was not general purpose. Xu et

al. (2023) improved that model to latency in vehicular

edge networks, but ignored the comparison of

performance among heterogeneous any devices and

regions. Zhang et al. (2024) introduced the concept

of collaborative perception with the vehicle-edge-

cloud paradigm, focusing on the possibility to obtain

an awareness of real-time traffic; however, they

evaluated their model on synthetic data and the

approach was not evaluated on real-world conditions.

Some studies concentrated on a static structure

instead of an adaptive one. Lin et al. (2021), they

proposed a fixed-resource allocation low-latency

framework that can be sub-efficient in the presence of

high-density traffic. Kumar and Goudar (2022) also

studied latency-optimized edge systems and

neglected energy consumption constraints.

Meanwhile, Huang et al. (2021) utilized 5G technolo-

gies to further enhance edge responsiveness, but

deployment may be difficult in rural regions as it is

dependent on an emergent infrastructure.

Mahmud et al. (2022) provided a survey about

dynamic resource management for edge computing

and does not focus on autonomous driving, which

creates a gap in terms of domain-specific insight.

Zhang et al. (2023) presented edge AI for vehicular

communication, but the complexity of models and

their size limit the scalability and inference time.

Tang et al. (2021) analyzed edge-based sensor data

processing and provided insights on latency

optimization with limited testing.

A recent study (Zhao and Chen 2024) proposed

adding the privacy-preserving feature of federated

learning to AV networks, although both the security

issues of AV and the problems of data

synchronization have not been sufficiently discussed.

Lei and Zhou (2023) studied deep learning in

dynamic task offloading at the edge, but the model

was not transparent and explainable. Du et al. (2022)

presented a V2V protocol based on edge resources,

however their protocol depends on static urban

patterns, limiting the flexibility.

Emergency-based models similar to Ranaweera

and Perera (2021) discuss the low-latency of AV

hazard signals only and cannot be simply extended to

the general driving scenario. Yu et al. (2021)

considered URLLC model assuming perfect network

conditions which are hardly faced in the real

practice. Qiu et al. (2023) presented lightweight

computation models but experiment only on lane-

keeping tasks, restricting its applicability.

Dynamic offloading for mobile edge computing

in AVs was proposed in Hassan and Singh (2022) but

safety-based task prioritization at the task-level was

not considered. Fan et al. (2021) presented a

hierarchical EFC model, without considering latency

among communication tiers. Kundu and Ghosh

(2024) provided a review of low-latency vehicular

systems, but did not introduce new architectures nor

provide empirical results.

Wei and Ren (2023) studied edge computing over

vehicle platooning, but did not generalize the results

for different traffic conditions. Finally, Liu et al.

(2024) presented a cooperative edge-based

vehicular-to-all (V2X) model, however it did not

consider network failure and dynamic

reconfiguration in disconnected domains.

Taken together, these works have suggested

significant advances on the integration of edge

computing and AVs, yet still suffer in some

dimensions such as field trial, dynamic adaptivity,

safety-aware priority, and energy-aware task

execution. These constraints are the motivation of this

work that seeks to design a reliable and context-

aware edge computing architecture to satisfy the

immediate and uncertain nature of AV ecosystems.

3 METHODOLOGY

In this paper, we employ a hybrid, context-sensitive

and context-driven approach to introduce, implement

and evaluate a state-of-the-art real-time edge

computing architecture for autonomous vehicles

(AVs) that can deploy in latency-sensitive

environments. The approach is decomposed into three

main stages: system design, contextual modeling and

experimental validation with real traffic conditions.

The main goal of the system will be to provide a

computationally efficient and safety-oriented

architecture with consideration to energy

consumption and the proximity of edge nodes for

reducing the communication delay, which will work

smartly according to the vehicular context.

System Architecture The system architecture is

composed of onboard sensors, V2X modules, andalso

road side edge servers in a cooperative manner. As

part of the edge layer there is also a light-weight

decision-making engine that is responsible for

processing real-time inputs from several AVs and

scheduling tasks according to its urgency and safety

relevance. The architecture includes a multi-level

scheduler that organizes incoming data packets in

order of urgency – such as detecting an obstacle,

proximity to pedestrians or braking commands – and

operates on high priority data in preference to the

A Contextâ

SAware and Energyâ

SEfﬁcient Edge Computing Framework for Lowâ

SLatency Communication in Autonomous Vehicles

with Realâ

SWorld Validation and Safetyâ

SCentric Task Prioritization

non-critical operations of infotainment or

environmental monitoring.

Context awareness is realized through constant

fusion of input data from an on-board GPS, LIDAR,

cameras and traffic signal (stereo vision) to enable the

edge engine to evaluate environmental factors like

traffic concentration, type of road, weather and

lighting conditions. These contextual factors drive

dynamic resource allocation, allowing the system to

promptly change both its computational and

networking tactics. It also includes a federated

learning module to enable decentralized knowledge

collaboration among edge nodes without

compromising data privacy, as well as to enhance

system intelligence without reliance on a centralized

facility. Table 1 represents the task prioritization

levels in the edge scheduler.

Table 1: Task Prioritization Levels in the Edge Scheduler.

Priority

Level

Task

Type

Example

Processing

Time Target

High

Safety

Critic

Obstacle

detection,

emergency

braking

< 20 ms

Medium

Drivin

Assist

ance

Lane

detection,

adaptive

cruise

< 50 ms

Low

Non-

Critic

al/Use

Servic

Infotainme

nt, map

updates

< 200 ms

For experimental evaluation, we simulate the

framework and implement it using some real traffic

datasets and vehicular trajectory traces, to evaluate

the performance of latency, task completion ratio,

energy consumption and communication reliability.

Performance comparisons with traditional cloud-

based and non-contextual edge frameworks are

provided. Effectiveness of the proposed model is

evaluated using metrics, such as end-to-end delay,

edge nodes response time, and accuracy of safety

response. The virtualization tools are considered in

the execution environment to emulate a vehicle-to-

edge interaction, taking place in the context of

vehicular networks, and the hardware emulation of

edge nodes allows one to analyze resource

consumption and scalability when the vehicular

density varies.

In doing so, we show how to integrate these three,

to yield a context-aware, latency optimized, and

safety centric edge computing system, providing

dramatic improvements in autonomy vehicle opera-

tion responsiveness and confidence in dynamic

environments. Figure 1 shows the proposed edge

computing framework for low-latency

communication in autonomous vehicles.

Figure 1: Proposed Edge Computing Framework for Low-

Latency Communication in Autonomous Vehicles.

4 RESULTS AND DISCUSSION

The simulation analysis of the proposed edge

computing schema shows the progress in the low-

latency communication and safety-critical

responsiveness with the autonomous vehicle

scenarios. Evaluation of the developed model against

traditional cloud-based architecture and the existing

non-contextual edge shows improvements between

both architectures using different performance

metrics which shows that the presented context-aware

edge model help in enhancing edge integration.

Perhaps most significantly is the large decrease

in end-to-end latency. Under the same traffic

conditions, the proposed system always kept the

communication delay less than 25 ms for safety-

critical traffic, which was much lower than those of

the cloud systems with often over 80 ms. This

decrease can be primarily attributed to edge-based

computation and smart scheduling: the most urgent

vehicle instructions can be handled immediately

without having to wait in line with less critical data

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

flows. Figure 2 and table 2 shows the average latency

comparison.

Figure 2: Average Latency Comparison.

Regarding the efficiency of task completion, the

context-aware scheduler will be able to dynamically

react to the condition of roads, density of traffic and

environment. The context-based prioritization of

tasks in the point process model resulted in enhanced

overall system responsiveness even at high loads,

such as urban intersections and multi-vehicle

interactions. The scheduler correctly classified high-

priority and low-priority events with 92% accuracy

temporarily filing time-critical tasks (such as

emergency braking or obstacle avoidance) until these

could be executed.

Table 2: Latency Comparison Across Architectures.

Architectu

re Type

Average

Latency

(ms)

Peak

Latency

(ms)

Task

Failure

Rate (%)

Cloud-

Based

130

7.8

Traditiona

l Edge

5.2

Proposed

Framewor

1.3

The analysis of another important dimension was

that of high energy-efficiency. The lightweight

processing models of the framework, as well as its

dynamic resource allocation policies, enable the

reduction in power consumption at the edge nodes by

28% with respect to the baseline edge systems, which

do not integrate energy-aware mechanisms. This

enhancement is particularly beneficial for being

deployed in infrastructure with scarce energy supply,

or vehicular applications with long-duration

continuous operation requirements.

In addition, the federated learning support enabled

improved system intelligence with time. During test

cycles, predictive accuracy of safety threats

progressed incrementally because around the edge

nodes learned from nearby experiences and shared

updated parameters without the need of centralized

training. Interestingly, this approach generalizes

better to semi-urban or foggy environments than

traditional approaches, for which higher error rates

are usually seen in these other scenarios.

Figure 3: Accuracy of Context-Aware Task Prioritization.

Although there are such enhancements, this

discussion also presents a few limitations. Although

the system proved to be effective under control and

semi-control settings, additional tests in extremely

unstructured rural terrains are required to confirm its

scalability and reliability in non-standard conditions.

Furthermore, as privacy-preserving federated

learning offered improvement in mitigating the risks

of data centralization, it still needs more studies to

face adversarial threats and data jigging at the edge

level. Figure 3 and table 3 shows the accuracy of

context-aware task prioritization.

Table 3: Accuracy of Context-Aware Task Prioritization.

Scenario

Detecte

Context

Correct

Task

Assigned

Accur

acy

(%)

Urban

Intersection –

Daylight

High

Traffic

Emergenc

Response

94.6

Highway –

Night

Low

Visibilit

Speed

Regulation

92.1

School Zone –

Rain

Wet

Surface

Braking

Optimizati

95.3

Semi-Urban

Road – Dense

Fog

Low

Visibilit

Sensor

Fusion

Alert

91.8

Tunnel Exit –

Daylight

Glare

Conditi

Vision

Recalibrati

93.2

In total, results confirm that a context-sensitive,

latency-optimized, and energy-efficient edge

computing framework is indeed able to fulfill the

strict requirements of real-time autonomous driving.

The discussion demonstrates that the incorporation

A Contextâ

SAware and Energyâ

SEfﬁcient Edge Computing Framework for Lowâ

SLatency Communication in Autonomous Vehicles

with Realâ

SWorld Validation and Safetyâ

SCentric Task Prioritization

of dynamic environmental awareness, safety-centric

task scheduling, and federated learning can enhance

low latency and performance, and effectively can

contribute the establishment of a scalable and

intelligent vehicular communication ecosystem

toward future intelligent transportation networks.

Figure 4 shows the average power usage of edge

system.

Figure 4: Average Power Usage of Edge Systems.

5 CONCLUSIONS

In this paper, we propose a new edge computing

architecture for AVs by focusing on low-latency

communication, context awareness, and safety-

critical task scheduling requirements. The proposed

system overcomes important drawbacks of the

current vehicular communication architectures by a

smart combination of real-time environment data

integration, lightweight processing and adaptive

scheduling. In contrast to conventional cloud-based

models, this paradigm moves computational

intelligence near data source, which can be beneficial

for reducing response time and for the timely

execution of life-critical decisions.

Results validate the beneficial of utilizing

contextual awareness and dynamic resource

management for responsive and efficient edge

systems in challenging driving scenarios. In addition

to the efficiency in terms of latency and energy

consumption, the model scales well with the traffic

scenario, thanks to federated learning techniques, that

enable distributed sharing of learned knowledge

without sacrificing data privacy.

By connecting the theoretical edge-computing

model with the practical requirements of AV

deployment, this work opens the door for developing

dependable, adaptive, and smart transport systems. It

confirms that the future of self-driving cars will be

based not only on fast computation, but on

processing information smartly and contextually at

the edge. Future research will engage enriched

framework’s robustness in heterogeneous networks,

increasing security for the edge layer, and the

validation on larger scale smart city infrastructures.

REFERENCES

Ahmed, M., & Kim, B. S. (2021). A hybrid edge-cloud

framework for latency reduction in autonomous driv-

ing. Sensors, 21(11), 3679.

https://doi.org/10.3390/s21113679

Alghamdi, T., & Baz, A. (2021). Edge computing for au-

tonomous vehicles: A survey on architectures and chal-

lenges. IEEE Access, 9, 133972–133989.

https://doi.org/10.1109/ACCESS.2021.3116403

Du, M., Wang, L., & He, Q. (2022). Edge-based V2V com-

munication protocol design for urban mobility. Com-

puter Communications, 187, 234–242.

https://doi.org/10.1016/j.comcom.2022.01.013

Fan, X., Liu, Q., & Chen, T. (2021). Enhancing AV safety

via edge-fog-cloud hierarchy. Future Internet, 13(8),

204. https://doi.org/10.3390/fi13080204

Hassan, S., & Singh, K. (2022). Offloading mechanisms for

time-critical autonomous vehicle tasks to MEC. Mobile

Networks and Applications, 27, 1023–1034.

https://doi.org/10.1007/s11036-021-01765-z

Huang, J., Feng, Z., & Yang, Y. (2021). 5G-enabled vehic-

ular edge computing for improved safety in AVs. IEEE

Internet of Things Journal, 8(8), 6790–6800.

https://doi.org/10.1109/JIOT.2020.3036223

Kumar, N., & Goudar, R. H. (2022). Latency-optimized

edge computing model for connected vehicle applica-

tions. Vehicular Communications, 34, 100443.

https://doi.org/10.1016/j.vehcom.2022.100443

Kundu, S., & Ghosh, T. (2024). A review on low-latency

vehicular communication techniques. IEEE Access, 12,

17852–17869. https://doi.org/10.1109/AC-

CESS.2024.3345312

Lei, H., & Zhou, F. (2023). Task migration for AVs in edge

networks: A deep learning approach. Neurocomputing,

528, 128139. https://doi.org/10.1016/j.neucom.2023.0

2.024

Lin, X., Wang, J., & Gao, L. (2021). Edge computing based

low-latency framework for autonomous driving sys-

tems. Journal of Systems Architecture, 117, 102118.

https://doi.org/10.1016/j.sysarc.2021.102118

Liu, Y., Pan, Z., & Tang, J. (2024). Cooperative decision-

making for AVs with edge-enabled V2X. IEEE Trans-

actions on Intelligent Transportation Systems. Advance

online publication.

https://doi.org/10.1109/TITS.2024.3356289

Mahmud, R., Koch, F., & Buyya, R. (2022). Dynamic re-

source management in edge computing for real-time au-

tonomous driving. ACM Computing Surveys, 55(1), 1–

34. https://doi.org/10.1145/3469028

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

Qiu, T., Wu, M., & Ma, J. (2023). Lightweight edge com-

putation for real-time driving scenarios. IEEE Transac-

tions on Mobile Computing, 22(4), 1675–1688.

https://doi.org/10.1109/TMC.2022.3161659

Rahman, M. A., & Mehedi, M. H. (2022). Low-latency

V2X communication for safety-critical applications us-

ing vehicular edge computing. Ad Hoc Networks, 125,

102731. https://doi.org/10.1016/j.adhoc.2021.102731

Ranaweera, D., & Perera, C. (2021). An edge-centric archi-

tecture for AV emergency communication. Journal of

Network and Computer Applications, 176, 102954.

https://doi.org/10.1016/j.jnca.2020.102954

Shen, Z., Liu, H., Zhang, L., & Wu, H. (2023). Cooperative

edge intelligence for connected autonomous vehicles:

Framework and challenges. IEEE Network, 37(2), 88–

94. https://doi.org/10.1109/MNET.123.2100545

Sun, Y., Guo, Y., Wang, L., & Yu, F. R. (2021). Edge com-

puting and caching for autonomous driving: A deep re-

inforcement learning approach. IEEE Wireless

Communications, 28(3), 94–100.

https://doi.org/10.1109/MWC.001.2000362

Tang, J., Zhang, Y., & Wu, J. (2021). Edge computing for

AV sensor data processing: A case study of latency and

safety. Sensors, 21(4), 1451.

https://doi.org/10.3390/s21041451

Wang, Z., Wang, S., Liu, Y., & Liu, A. (2022). Delay-aware

task offloading for autonomous driving in vehicular

edge computing. IEEE Transactions on Vehicular

Technology, 71(2), 1228–1241.

https://doi.org/10.1109/TVT.2021.3134821

Wei, X., & Ren, J. (2023). Edge computing for AV platoon-

ing control under latency constraints. Sensors, 23(6),

3157. https://doi.org/10.3390/s23063157

Xu, L., Zheng, K., & Wang, H. (2023). Edge computing for

safety and efficiency in autonomous vehicular net-

works. Computer Networks, 229, 109859.

https://doi.org/10.1016/j.comnet.2023.109859

Yu, R., Xie, S., & Zhang, Y. (2021). Towards ultra-reliable

low-latency communication for AVs using edge infra-

structure. IEEE Communications Magazine, 59(1), 98–

104. https://doi.org/10.1109/MCOM.001.2000236

Zhang, W., Li, X., & Wu, C. (2023). Real-time vehicular

communication using edge AI. IEEE Transactions on

Intelligent Transportation Systems, 24(1), 563–573.

https://doi.org/10.1109/TITS.2022.3140174

Zhang, Y., Wang, J., & Liu, Y. (2024). Vehicle-edge-cloud

collaboration for real-time traffic perception and deci-

sion-making. Future Generation Computer Systems,

145, 4052. https://doi.org/10.1016/j.future.2023.09.01

Zhao, L., & Chen, H. (2024). Federated learning and edge

computing integration in intelligent vehicles. Infor-

mation Fusion, 92, 139–152.

https://doi.org/10.1016/j.inffus.2023.07.005

A Contextâ

SAware and Energyâ

SEfﬁcient Edge Computing Framework for Lowâ

SLatency Communication in Autonomous Vehicles

with Realâ

SWorld Validation and Safetyâ

SCentric Task Prioritization