5. mmWave Localization Perspective

Why mmWave Helps Localization

Millimeter-wave signals usually refer to frequencies from about 30 GHz to 300 GHz. Their wavelengths are short, so large antenna arrays can fit into compact hardware. Large arrays provide narrow beams and high angular resolution.

High bandwidth also improves delay resolution:

$$\Delta\tau \approx \frac{1}{B} \quad\Longrightarrow\quad \Delta d \approx \frac{c}{B}$$

so 400 MHz resolves paths about 75 cm apart, and 2 GHz about 15 cm.

This makes mmWave useful for both communication and positioning.

Why mmWave localises better: resolution at a glance

1. Formula and symbols

$$\Delta d\approx c/B,\quad\Delta\theta\approx2/M,\quad\Delta x\approx r\Delta\theta$$

• $B$: bandwidth
• $M$: elements across the array
• $r$: range
• $\Delta d$: one-way path-length resolution
• $\Delta \theta$: angular resolution in radians
• $\Delta x$: cross-range separation

Monostatic round-trip range instead uses c/(2B).

Your bandwidth

800 MHz

Your elements per side

64

User distance

20 m

2. Work with the current inputs

$$\Delta d \approx \frac{c}{B} = \frac{3\times10^{8}}{800\times10^{6}} = 37.5\ \text{cm}, \qquad \Delta\theta \approx \frac{2}{M} = \frac{2}{64}\ \text{rad} = 1.79^\circ, \qquad \Delta x \approx r\,\Delta\theta = 62.5\ \text{cm}$$

3. Result

Your range resolution37.5 cm

Your beamwidth1.79°

Cross-range at 20 m62.5 cm

Resolution cell37.5 cm × 62.5 cm

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.

Why mmWave Needs RIS

The same properties that make mmWave powerful also make it fragile:

• Severe blockage.
• Higher penetration loss.
• Narrow beams that require alignment.
• Sensitivity to mobility and orientation.

RIS can provide a controllable reflected path when direct mmWave links fail.

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.

Beam Training

Before data transmission, mmWave systems must find good beams. With RIS, this becomes a joint problem:

• BS transmit beam.
• RIS phase pattern.
• UE receive beam.

Naively testing every combination is too slow. Practical methods use hierarchical beam codebooks, compressive sensing, machine learning, or position-aided beam search.

Beam training: can we finish before the channel changes?

1. Formula and symbols

$$T_{\rm exh}=N_BN_RN_U T_s$$

$$T_{\rm hier}\approx2(\log_2N_B+\log_2N_R+\log_2N_U)T_s$$

$$T_c\approx0.423/f_D,\quad f_D=vf_c/c$$

• $N_B$: Number of BS candidate beams
• $N_R$: Number of RIS candidate beams
• $N_U$: Number of UE candidate beams
• $T_s$: time per test
• $T_c$: coherence estimate
• $v$: speed
• $f_c$: carrier
• $c$: wave speed

Hierarchical search uses this idealized binary-search count.

BS beams

64

RIS patterns

1,024

UE beams

4

User speed

30 km/h

2. Work with the current inputs

$$\begin{aligned} T_{\text{exh}} &= N_{\text{BS}}N_{\text{RIS}}N_{\text{UE}}\,T_s = 262{,}144\times 8.9\ \mu\text{s} \\ T_{\text{hier}} & \approx 2\left(\log_2 N_{\text{BS}} + \log_2 N_{\text{RIS}} + \log_2 N_{\text{UE}}\right)T_s \\ T_c &\approx \frac{0.423}{f_D},\quad f_D = \frac{v f_c}{c} = 778\ \text{Hz} \end{aligned}$$

3. Result

Exhaustive262,144 probes2.34 s

Hierarchical36 probes321 µs

Coherence time543 µsDoppler 778 Hz

Exhaustive fits?No4300.9× the coherence time

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.

Near-Field Effects

For large RIS apertures, the user may be in the near-field region. In this region, beams focus on points rather than only directions.

This enables:

• Range-angle localization.
• User-specific focusing.
• More spatial degrees of freedom.
• Improved separation of nearby users or targets.

But it also requires more accurate channel models.

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.

RIS Orientation and Placement

RIS localization performance depends strongly on where the RIS is placed:

• A wall-mounted RIS can help a blocked street or corridor.
• A ceiling RIS can improve indoor positioning.
• Multiple RIS panels can improve geometry.
• Poorly placed RIS panels create weak or ambiguous measurements.

RIS orientation matters because a surface has a limited useful angular sector. A panel facing the wrong direction may provide little energy.

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.

Main Research Problems

• Fast RIS-aided beam training.
• Sparse channel estimation with low pilot overhead.
• Near-field spherical-wave modeling.
• Joint communication-rate and localization-error optimization.
• Learning-based fingerprinting using RIS beam patterns.
• Robustness under blockage and mobility.

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.

Takeaway

• mmWave contribution: wide bandwidth and large electrical apertures can improve delay/angle resolution.
• RIS contribution: a controlled alternate path when the direct route is blocked.
• Cost: beam training, channel estimation, and near-field modelling must be accurate.

References for this section3

• K. Li et al., "Localization in Reconfigurable Intelligent Surface Aided mmWave Systems", IEEE Transactions on Vehicular Technology, 2024.
• X. Mu et al., "Reconfigurable Intelligent Surface-Aided Near-Field Communications for 6G", IEEE Vehicular Technology Magazine, 2024.
• K. M. Faisal and W. Choi, "Machine Learning Approaches for Reconfigurable Intelligent Surfaces", IEEE Access, 2022.