🔢 mathematics

Environment-Aware Near-Field Channel Estimation Leveraging CKM and ISAC

This paper proposes an environment-aware near-field channel estimation framework for ISAC systems with extremely large-scale antenna arrays that leverages a novel virtual object map (VOM) for static multipath characterization and a sensing-assisted protocol for dynamic target detection to significantly improve estimation accuracy and achievable rates.

Original authors: Yuan Guo, Yilong Chen, Zixiang Ren, Jie Xu

Published 2026-04-07

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Yuan Guo, Yilong Chen, Zixiang Ren, Jie Xu

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

The Big Picture: Navigating a Foggy City with a GPS and a Radar

Imagine you are trying to drive a car (the User) from a specific spot to a massive, high-tech lighthouse (the Base Station) in a dense, foggy city. The lighthouse has thousands of tiny lights on it (an Extremely Large Antenna Array) that need to shine a perfect, focused beam of light directly into your car's windshield so you can see clearly.

The problem? The city is full of obstacles: tall buildings, walls, and moving cars. The light doesn't just travel in a straight line; it bounces off buildings, scatters, and reflects. This is called the Near-Field Channel. To aim the light perfectly, the lighthouse needs to know exactly where every bounce happens.

Traditionally, the lighthouse would have to shout a test signal ("Hello?") and wait for you to shout back. But in a city with thousands of lights and complex bounces, shouting back takes too much time and energy, and you might miss a moving car.

This paper proposes a smarter way: The Lighthouse uses a "Digital Map" and its own "Radar" to guess the path before it even shouts.

The Three Key Ingredients

1. The "Virtual Object Map" (VOM) – The Digital City Map

Instead of trying to memorize every brick in every building, the lighthouse uses a Channel Knowledge Map (CKM).

The Analogy: Imagine you have a digital map of the city that doesn't just show buildings, but shows "Virtual Bounce Points." These are specific spots on walls or corners where light usually bounces.
How it works: The lighthouse knows, "If the car is at this street corner, the light will likely bounce off that specific wall." This is the Static part of the problem. The buildings don't move, so the map is always right.
The Paper's Twist: They call this the Virtual Object Map (VOM). It's like a cheat sheet that tells the lighthouse, "Don't look everywhere; just focus on these 5 specific bounce points for this user." This saves a huge amount of time.

2. Integrated Sensing and Communication (ISAC) – The Lighthouse's Radar

While the lighthouse is trying to talk to you, it is also acting like a radar.

The Analogy: Think of the lighthouse as a bat. It sends out a sound (the pilot signal). Most of the sound bounces off the buildings (the static stuff we already know from the map). But some sound hits a moving bird (a Dynamic Target or a moving car).
How it works: The lighthouse listens to the echoes. It uses the "Digital Map" (VOM) to ignore the echoes from the buildings (static clutter). What's left? The echoes from the moving bird.
The Paper's Twist: By filtering out the "noise" of the buildings, the lighthouse can instantly see where the moving objects are and how they are changing the path of the light. This is the Dynamic part.

3. The "Quantized Feedback" – The Shorthand Note

You (the user) receive the lighthouse's test signal. You can't shout back a detailed description of the whole city.

The Analogy: You have a pre-agreed list of codes (a codebook). You look at the signal, pick the code that matches best, and send back a tiny number (like "Code #42").
How it works: The lighthouse receives "Code #42," looks it up, and says, "Ah, that means the signal is coming from direction X with strength Y."
The Paper's Twist: Because the lighthouse already has the Map (VOM) and the Radar (Sensing) data, it doesn't need a long, detailed description from you. It only needs that tiny "Code #42" to fill in the final missing pieces.

How They Put It All Together (The Recipe)

The paper proposes a three-step recipe to aim the beam perfectly:

Consult the Map (VOM): The lighthouse looks at your location and pulls up the "Virtual Object Map." It knows exactly where the static bounces (walls) are. It builds a "skeleton" of the signal path based on this.
Scan for Movement (Sensing): The lighthouse sends out a signal and listens to the echoes. It uses the Map to ignore the walls. It isolates the echoes from moving objects (like a car driving by) and figures out how they are distorting the signal.
Fill in the Blanks (Feedback): You send back your tiny "Code #42." The lighthouse combines its Map, its Radar scan, and your code to calculate the exact strength of the signal.

The Result: Instead of guessing the whole path from scratch (which is slow and hard), the lighthouse just has to adjust a few numbers. It's like finishing a puzzle where you already have the border and the sky; you only need to place the few missing pieces.

Why Does This Matter? (The Results)

The paper ran computer simulations to prove this works better than old methods.

Accuracy: The new method is much more accurate at aiming the beam. It's like hitting a bullseye in a dark room using a flashlight, whereas the old method was like throwing darts in the dark.
Speed/Efficiency: Because the lighthouse already knows the "static" part of the city, it doesn't need to shout as many test signals. This saves battery and frees up space for actual data (like streaming video).
The "Short Pilot" Win: The biggest win is when you have very little time to send test signals. The new method shines here because it relies on the "Map" and "Radar" to do the heavy lifting, rather than relying on you to shout back a long message.

Summary in One Sentence

This paper teaches 6G towers to use a pre-made digital map of the city and their own built-in radar to guess where the signal bounces, so they can aim their beams perfectly with very little help from your phone.

1. Problem Statement

The paper addresses the challenge of acquiring accurate Channel State Information (CSI) for Extremely Large-Scale Antenna Arrays (ELAAs) operating in the near-field regime for 6G Integrated Sensing and Communication (ISAC) systems.

The Challenge: Conventional channel estimation relies on real-time pilot-based training. For high-dimensional near-field channels, this incurs prohibitive pilot overhead and fails to exploit available environmental information.
The Context: Wireless channels consist of static components (determined by persistent structures like buildings) and dynamic components (caused by moving targets). Existing methods often treat the channel as an unstructured whole, ignoring the separability of these components and the potential to use environmental priors.

2. Methodology

The authors propose a novel Environment-Aware Near-Field Channel Estimation Framework that jointly leverages Channel Knowledge Maps (CKM) and ISAC capabilities.

A. System Model

Scenario: A downlink scenario where an ISAC-enabled Base Station (BS) with an ELAA serves a single-antenna User Equipment (UE).
Channel Decomposition: The channel is modeled as a superposition of Multipath Components (MPCs) arising from:
- Type-1 EOs: Static compact objects (reflection/scattering).
- Type-2 EOs: Static large smooth surfaces (specular reflection).
- Sensing Targets (STs): Dynamic moving objects.
Signal Model: Both the downlink communication channel and the monostatic sensing channel are modeled using near-field array responses associated with these interaction points.

B. Key Innovation: Virtual Object Map (VOM)

Instead of storing raw physical environmental data, the paper introduces a Virtual Object Map (VOM) as a novel CKM representation:

Concept: A finite library of "virtual objects" representing the dominant primary interaction locations (BS-visible points) induced by static Type-1 and Type-2 EOs.
Function: For a given UE location, the VOM maps the location to a specific index set of the top- $J$ dominant virtual objects. This provides a location-dependent, reusable prior for the static channel component.
Construction: Built offline using ray-tracing and spatial clustering to identify dominant static paths.

C. Joint Estimation Protocol

The framework operates in a time-slotted coherence block:

Training Slot:
- The BS transmits downlink pilots ( $Z$ ).
- The UE receives pilots, quantizes the observation, and feeds back the index ( $\hat{b}$ ).
- Simultaneously, the BS collects monostatic echoes ( $E$ ) from the environment.
Dynamic Information Extraction (Sensing):
- The BS uses the VOM (specifically the mapping for the BS location) to identify the subspace of static clutter (Type-1/2 EOs).
- It projects the echo signal onto the orthogonal complement of this static subspace to suppress static clutter.
- The residual signal is processed via Singular Value Decomposition (SVD) to extract a low-dimensional dynamic subspace representing the moving targets (STs).
Channel Estimation:
- The downlink channel is approximated as a linear combination of:
  - Static Basis: Retrieved from the VOM (associated with the UE location).
  - Dynamic Basis: Extracted from the sensing echoes.
- The BS solves a regularized least-squares problem to estimate the complex coefficients for both the static and dynamic components using the quantized pilot feedback.

3. Key Contributions

Virtual Object Map (VOM): Proposed a new CKM representation that stores locations of virtual interaction points rather than raw physical objects, enabling efficient construction of near-field array response bases for both static and sensing channels.
Joint VOM-and-Sensing Framework: Developed a protocol that simultaneously utilizes VOM priors for static components and real-time ISAC sensing for dynamic components, significantly reducing the dimensionality of the estimation problem.
Clutter Suppression & Subspace Extraction: Designed a method to suppress static environmental echoes in sensing data to isolate dynamic target information, which is then used to refine channel estimation.
Structured Estimation: Formulated the channel estimation as a structured problem (Static + Dynamic) rather than an unstructured search, allowing for accurate recovery with limited pilot overhead.

4. Simulation Results

The authors evaluated the scheme using a 64-antenna ULA at 2.4 GHz.

Metrics: Normalized Mean Squared Error (NMSE) and Achievable Downlink Rate.
Benchmarks: Compared against:
- VOM-only (no dynamic sensing).
- Conventional (no VOM, no sensing).
Findings:
- Accuracy: The proposed joint scheme achieved the lowest NMSE across all pilot lengths ( $T_p$ ), significantly outperforming benchmarks. The gap was most pronounced in the short-pilot regime, proving effectiveness when resources are scarce.
- Throughput: The proposed scheme achieved the highest achievable downlink rate, closely approaching the "Perfect CSI" upper bound.
- Synergy: The VOM-only scheme outperformed the conventional scheme, confirming the value of static priors. However, adding dynamic sensing provided further gains, demonstrating the necessity of tracking dynamic targets.

5. Significance

Overhead Reduction: By exploiting environmental priors (VOM) and sensing, the framework drastically reduces the required pilot overhead for high-dimensional near-field channels, a critical bottleneck for 6G ELAA systems.
ISAC Synergy: It demonstrates a practical application of ISAC where sensing data is not just for target detection but is directly integrated to enhance communication performance (channel estimation).
Scalability: The structured approach (separating static and dynamic components) makes channel estimation feasible for extremely large arrays where traditional methods would fail due to the "curse of dimensionality."
Future Outlook: The work paves the way for adaptive environment-aware communications where maps are updated online, further enhancing robustness in dynamic 6G networks.