⚡ electrical engineering

Near-Field Integrated Sensing, Computing and Semantic Communication in Digital Twin-Assisted Vehicular Networks

This paper proposes a near-field integrated sensing, computing, and semantic communication (ISCSC) framework for digital twin-assisted vehicular networks that leverages MU-MIMO and mmWave radar to jointly optimize vehicle tracking and data transmission, achieving a 20% rate improvement while maintaining sensing accuracy under constrained resources.

Original authors: Yinchao Yang, Yahao Ding, Jiaxiang Wang, Zhaohui Yang, Chen Zhu, Zhaoyang Zhang, Dusit Niyato, Mohammad Shikh-Bahaei

Published 2026-04-08

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yinchao Yang, Yahao Ding, Jiaxiang Wang, Zhaohui Yang, Chen Zhu, Zhaoyang Zhang, Dusit Niyato, Mohammad Shikh-Bahaei

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a futuristic highway where every car, every traffic light, and every road sign is connected to a massive, super-smart "Digital Twin." This Digital Twin is like a perfect, real-time video game copy of the real world. It knows exactly where every car is, how fast it's going, and where it's heading, allowing the system to prevent accidents and manage traffic flow instantly.

However, keeping this Digital Twin updated is incredibly hard. It requires three things to happen at the exact same time, often competing for the same resources:

Sensing: The system needs to "see" the cars (like a radar) to know where they are.
Talking: It needs to send messages to the cars (like a Wi-Fi signal) to tell them what to do.
Thinking: It needs to process all this data to update the Digital Twin model.

The problem is that in the future, these systems will use very high-frequency waves (like 5G or 6G). At these frequencies, the waves don't travel in straight, flat lines anymore; they curve like ripples in a pond. This is called the "Near-Field" effect. Most existing systems assume waves are flat, which causes them to get confused and make mistakes when cars are close to the roadside units.

The Solution: The "Smart Trio" Framework

This paper proposes a new framework called ISCSC (Integrated Sensing, Computing, and Semantic Communication). Think of it as a highly efficient, three-in-one team working at every roadside unit (RSU).

Here is how it works, using simple analogies:

1. The "Smart Radar" (Sensing)

Instead of just sending out a generic radar ping, the roadside unit uses a massive array of antennas (like a giant spider web of sensors). Because the cars are close, the system accounts for the curving "Near-Field" waves.

The Analogy: Imagine trying to hear a whisper in a crowded room. Old systems use a simple ear (linear waves) and might miss the whisper if it's close. This new system uses a "super-ear" that understands exactly how sound curves in a small room, allowing it to pinpoint the whisper's location with extreme precision.
The Result: It tracks every car's speed and position perfectly, even in a traffic jam.

2. The "Smart Messenger" (Semantic Communication)

Traditional communication is like sending a fax of a whole book just to tell someone, "The book is on the table." It sends everything, including redundant words.
This new system uses Semantic Communication. It's like sending a text message that just says, "Book on table."

The Analogy: Imagine you and your friend share a secret codebook. Instead of sending a 10-page document, you just send the code "A-1," and your friend knows exactly what you mean because they have the context.
The Benefit: This saves a massive amount of data and time. The roadside unit only sends the meaning of the message, not the raw data. This frees up bandwidth so the system can talk to more cars simultaneously.

3. The "Smart Brain" (Computing & Digital Twin)

The roadside unit has to build the Digital Twin model. If the radar makes a small mistake (e.g., thinks a car is 1 meter away when it's actually 1.1 meters), the computer has to work harder to correct that error in the model.

The Analogy: Think of the Digital Twin as a jigsaw puzzle. If the pieces (sensor data) are slightly blurry, you have to squint and think harder (use more CPU power) to fit them together. If the pieces are sharp, you can assemble the puzzle quickly.
The Innovation: The authors created a formula that links accuracy to effort. If the sensing is less accurate, the system automatically knows it needs to allocate more computer power to fix the model. This prevents the system from crashing under the load.

How They Work Together: The "Hybrid Heuristic"

The hardest part is deciding which roadside unit talks to which car, and how much power to use for sensing vs. talking. It's like a traffic controller trying to assign pilots to planes in a storm.

The authors developed a Hybrid Heuristic Algorithm.

The Analogy: Imagine a greedy algorithm is like a person who always picks the closest parking spot without looking at the whole lot. It's fast but often leads to a bad spot.
The new method starts with the "greedy" approach (picking the closest unit) but then uses a "Simulated Annealing" technique (like a smart explorer who occasionally tries a slightly further spot to see if it's better). It tests different combinations until it finds the perfect balance where no car is left waiting, and no computer is overheating.

The Results: Why It Matters

When they tested this system in simulations:

Speed: It increased the data transmission rate by 20% compared to older methods.
Accuracy: It kept the sensing accuracy just as high as the best existing systems, even though it was sending less data.
Efficiency: It managed the computer power so well that the Digital Twin stayed accurate without needing supercomputers.

In a Nutshell

This paper presents a blueprint for the future of smart highways. By combining curved-wave sensing, smart messaging (sending only what matters), and adaptive computing, the system creates a perfect Digital Twin of the road. It ensures that cars can talk to each other and the infrastructure instantly and accurately, making our future roads safer, faster, and less congested. It's the difference between a chaotic traffic jam and a perfectly choreographed dance.

1. Problem Statement

The paper addresses the critical challenges in deploying Digital Twin (DT) technology within future 6G vehicular networks. While DTs offer real-time, high-fidelity virtual representations of physical vehicles and infrastructure, their implementation faces three major bottlenecks:

Near-Field (NF) Effects: The use of high-frequency bands (mmWave/THz) and large antenna arrays in 6G creates strong near-field effects (spherical wave propagation) within the Rayleigh distance. Traditional far-field planar wave models are insufficient, leading to inaccurate sensing and communication.
Resource Competition: DT-assisted networks require simultaneous execution of Integrated Sensing and Communication (ISAC), Semantic Communication (SC), and Computing (for DT modeling). These functions compete for limited power, bandwidth, and computational resources.
Data Overhead: Conventional communication transmits redundant data, causing latency and bandwidth inefficiency. DTs require massive, real-time data updates, exacerbating these issues.

The core problem is how to jointly optimize sensing accuracy, semantic transmission rates, and computational resource allocation in a multi-user, near-field environment to maintain a high-fidelity Digital Twin without violating latency or power constraints.

2. Methodology

The authors propose a novel Integrated Sensing, Computing, and Semantic Communication (ISCSC) framework tailored for DT-assisted vehicular networks operating in the near-field regime.

A. System Model

Architecture: A Multi-User Multiple-Input Multiple-Output (MU-MIMO) system where multiple Roadside Units (RSUs) serve multiple vehicles.
Near-Field Modeling: Utilizes a Non-Uniform Spherical Wave (NUSW) channel model to accurately capture the spherical wavefronts and path-loss variations inherent in the near-field region (up to ~300m for 50 GHz).
Dual Functionality: Each RSU simultaneously:
1. Senses: Detects and tracks all vehicles in its coverage area using mmWave radar to update the DT.
2. Communicates: Serves a subset of vehicles using semantic communication to transmit control commands and updates.

B. Key Components

Semantic Communication (SC):
- Instead of transmitting raw bits, the system extracts semantic information (meaning) using an encoder (e.g., Transformer).
- A Semantic Extraction Ratio ( $\rho$ ) is optimized to balance the trade-off between transmission rate and the computational power required for extraction.
Sensing & Tracking:
- Particle Filtering (PF): The paper employs Particle Filtering (instead of EKF/UKF) at the RSUs to estimate vehicle states (angle, distance, velocity, path loss). PF is chosen for its robustness in handling non-linear and non-Gaussian noise common in dynamic vehicular environments.
- Sensing Accuracy: Measured using the Cramér-Rao Bound (CRB) for angle and distance estimation.
Computing for Digital Twin:
- A novel computational model links sensing accuracy to computing load.
- Key Insight: If sensing accuracy degrades (higher CRB), the DT model requires more computational effort to correct errors. The paper formulates the extra latency ( $\Delta T$ ) as a function of the Root CRB (RCRB) of distance and angle.
Optimization Framework:
- Objective: Minimize a weighted sum of negative semantic rate (maximize rate) and CRB (maximize accuracy), subject to power, latency, and CPU frequency constraints.
- Algorithm Design:
  - Vehicle Assignment: A Hybrid Heuristic (HH) algorithm combining a Greedy approach (nearest RSU) and Simulated Annealing (SA) to solve the non-convex vehicle-to-RSU assignment problem.
  - Resource Allocation: An Alternating Optimization (AO) approach is used to iteratively solve for beamforming matrices, semantic extraction ratios, and CPU frequencies.
  - Convexification: Non-convex constraints (rate and CRB) are transformed into convex forms using Lemma 1 (auxiliary variables) and Lemma 2 (low-SNR approximation).

3. Key Contributions

First Integrated ISCSC Framework for NF-DT: The paper introduces the first framework that unifies Near-Field ISAC, Semantic Communication, and Digital Twin computing, explicitly modeling the spherical wave effects of 6G.
Sensing-Computing Coupling Model: It proposes a novel mathematical relationship where the computational workload for DT modeling is directly dependent on sensing accuracy (RCRB). This allows the system to dynamically allocate CPU resources based on real-time sensing quality.
Hybrid Heuristic Algorithm: Development of a HH algorithm (Greedy + Simulated Annealing) to efficiently solve the NP-hard vehicle assignment problem, preventing sub-optimal local solutions common in greedy approaches.
Joint Optimization: A comprehensive solution that jointly optimizes beamforming, semantic extraction ratios, and computing resources, demonstrating that semantic communication can improve transmission rates without sacrificing sensing accuracy.

4. Numerical Results

Simulations were conducted with parameters typical of 6G vehicular networks (50 GHz, 310 transmit antennas, 3 receive antennas).

Tracking Performance:
- The proposed Particle Filter (PF) significantly outperformed Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) in terms of Root Mean Square Error (RMSE) for angle, distance, and velocity, especially as vehicle density increased.
- PF with 500 particles achieved a favorable trade-off between accuracy and computation time (< 2ms).
Semantic Rate vs. Vehicle Density:
- The proposed Hybrid Heuristic (HH) algorithm maintained a high semantic transmission rate (approx. 2.8 bps/Hz at 20 vehicles) compared to greedy algorithms, which dropped below 1.0 bps/Hz.
- The ISCSC framework achieved a 20% improvement in transmission rate compared to existing ISAC schemes while maintaining sensing accuracy.
Sensing Accuracy (CRB):
- The HH algorithm maintained low CRB for angle and distance even under high traffic congestion.
- The "Non-semantic" version of the HH algorithm showed similar sensing accuracy to the semantic version, proving that semantic extraction does not degrade sensing performance.
Computing Latency:
- Results confirmed that ignoring sensing errors (using a static computing model) leads to underestimating CPU requirements. The proposed model showed that higher sensing errors require significantly higher CPU frequencies to meet latency constraints.

5. Significance and Conclusion

This paper provides a foundational blueprint for 6G-enabled Digital Twin vehicular networks. Its significance lies in:

Bridging the Gap: It moves beyond theoretical ISAC by integrating the computational demands of Digital Twins and the efficiency of Semantic Communication.
Practical Near-Field Modeling: By addressing the near-field effects, it offers a realistic solution for the high-frequency, large-array deployments expected in 6G.
Resource Efficiency: The framework demonstrates that by intelligently managing the trade-offs between sensing, communication, and computing, it is possible to achieve high-fidelity DTs without overwhelming network resources.

The authors conclude that their ISCSC framework is a promising direction for future intelligent transportation systems, offering superior performance in dynamic, resource-constrained environments compared to existing state-of-the-art methods. Future work suggests exploring Multi-Access Edge Computing (MEC) for offloading and multi-modal data fusion.