Sensing Assisted Localization for ISAC
Introduction
Localization in cellular networks has evolved from Cell-ID and OTDOA in LTE to multi-RTT, DL-AoD, UL-AoA, and NR-PRS-based methods in 5G NR. These methods rely on dedicated positioning reference signals (PRS) and UE measurement reporting, achieving sub-meter accuracy under favorable conditions. However, PRS overhead, NLOS bias, and multipath errors remain significant challenges.
Sensing-assisted localization augments or replaces PRS-based methods by using radar-like sensing at the gNB to directly estimate UE (or target) positions. This approach does not require UE cooperation or dedicated PRS resources, and naturally handles NLOS environments by exploiting multipath rather than treating it as error.
Localization Approaches Comparison
┌──────────────────────────────────────────────────────────────────────┐
│ │
│ PRS-Based (Conventional): │
│ ┌──────┐ PRS ┌──────┐ Measure ┌──────┐ Report ┌──────────┐ │
│ │ gNB │───────▶│ UE │──────────▶│ UE │────────▶│ Location │ │
│ │ │ │ │ (ToA,AoA) │ │ │ Server │ │
│ └──────┘ └──────┘ └──────┘ └──────────┘ │
│ │
│ Sensing-Based: │
│ ┌──────┐ Sense ┌───────────┐ Extract ┌──────────────────────┐ │
│ │ gNB │────────▶│ Echo │──────────▶│ Position Estimate │ │
│ │ │ echo │ Process │ (R,θ,v) │ (No UE cooperation) │ │
│ └──────┘ └───────────┘ └──────────────────────┘ │
│ │
└──────────────────────────────────────────────────────────────────────┘
Key Concepts
Monostatic Range-Angle Localization
In monostatic sensing, the gNB transmits a sensing signal and receives the echo reflected from the target. The target’s position is estimated from:
Range \(R = \\frac{c \\cdot \\tau}{2}\), where \(\\tau\) is the round-trip delay and \(c\) is the speed of light.
Azimuth angle \(\\phi\) from AoA estimation using the antenna array.
Elevation angle \(\\theta\) (for 3D localization) using a 2D planar array.
Monostatic Range-Angle Localization
┌──────────────────────────────────────────────────────────────────┐
│ │
│ Target (UE/Object) │
│ ● │
│ ╱ ╲ │
│ R ╱ ╲ R │
│ (range)╱ ╲(range) │
│ ╱ θ ╲ │
│ ╱ (angle) ╲ │
│ ╱ ╲ │
│ ● gNB (Tx + Rx) │
│ │
│ Position = (R·cos(θ), R·sin(θ)) │
│ │
└──────────────────────────────────────────────────────────────────┘
Bistatic and Multi-Static Localization
When multiple gNBs cooperate for sensing, the localization accuracy improves through geometric diversity:
Bistatic: One gNB transmits, another receives the echo. The target lies on an ellipse with the two gNBs as foci.
Multi-static: Multiple transmitter-receiver pairs. Each pair defines an ellipse; the intersection gives the target position.
Multi-Static Localization Geometry
┌──────────────────────────────────────────────────────────────────┐
│ │
│ gNB₁ (Tx) Target ● gNB₂ (Rx) │
│ ●─ ─ ─ ─ ─ ─ ─ ─ ╱ ╲─ ─ ─ ─ ─ ─ ─ ─● │
│ ╲ ╱ ╲ ╱ │
│ ╲ Ellipse 1╱ ╲ Ellipse 2 ╱ │
│ ╲ ╱ ╲ ╱ │
│ ╲ ╱ ╲ ╱ │
│ ╲ ╱ ╲ ╱ │
│ ╲╱ Intersection ╲╱ │
│ = Target position │
│ │
│ gNB₃ (Tx/Rx) │
│ ●─ ─ ─ ─ ─ Ellipse 3 ─ ─ ─ ─ │
│ │
└──────────────────────────────────────────────────────────────────┘
NLOS Localization via Multipath Exploitation
Unlike PRS-based methods that suffer from NLOS bias, sensing can exploit multipath reflections for localization:
The gNB detects multiple echo paths with different delays and angles.
Each path corresponds to a reflection from a known or mappable surface (wall, building).
Using the geometry of the reflections, the target position can be estimated even when the direct (LOS) path is blocked.
This technique requires a pre-built or continuously updated environment map from sensing data.
Sensing-Derived Location for Communication Enhancement
Once the UE’s position is estimated via sensing, this location information improves communication in several ways:
Location-based beamforming: Direct beam steering towards the estimated position.
Location-based scheduling: Grouping co-located UEs for MU-MIMO spatial multiplexing.
Location-based handover: Triggering handover based on position rather than signal strength.
Location-based power control: Adjusting transmit power based on path loss predicted from distance.
Benefits
Benefit |
Description |
|---|---|
No dedicated PRS overhead |
Localization from sensing echoes avoids allocating PRS resources. |
No UE cooperation needed |
The gNB can localize targets without UE measurement reporting. |
NLOS robustness |
Multipath exploitation enables localization even without LOS. |
Continuous tracking |
Sensing provides continuous position updates without periodic reporting. |
Challenges
Range and angle resolution: Localization accuracy depends on bandwidth (range resolution = \(c / (2B)\)) and array aperture (angular resolution). Narrow bandwidth or small arrays limit accuracy.
Target association: In multi-target environments, associating sensing-detected positions with specific UEs requires additional information (e.g., timing advance correlation).
3D localization: Elevation angle estimation requires 2D planar antenna arrays with sufficient vertical aperture, which may not be available at all gNBs.
Environment map maintenance: Multipath-based NLOS localization requires an up-to-date map of the surrounding reflectors, adding computational and storage overhead.