Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Over-the-Air Fusion of Sparse Spatial Features for Integrated Sensing and Edge AI over Broadband Channels

Zhiyan Liu, Qiao Lan, Kaibin Huang

TL;DR

A novel framework, called Spatial Over-the-Air Fusion (Spatial AirFusion), which exploits radio waveform superposition to aggregate spatially sparse features over the air and thereby enables simultaneous access and optimally solving the resultant mixed-integer problem of Voxel-Carrier Pairing and Power Allocation (VoCa-PPA).

Abstract

The 6G mobile networks feature two new usage scenarios -- distributed sensing and edge artificial intelligence (AI). Their natural integration, termed integrated sensing and edge AI (ISEA), promises to create a platform that enables intelligent environment perception for wide-ranging applications. A basic operation in ISEA is for a fusion center to acquire and fuse features of spatial sensing data distributed at many edge devices (known as agents), which is confronted by a communication bottleneck due to multiple access over hostile wireless channels. To address this issue, we propose a novel framework, called Spatial Over-the-Air Fusion (Spatial AirFusion), which exploits radio waveform superposition to aggregate spatially sparse features over the air and thereby enables simultaneous access. The framework supports simultaneous aggregation over multiple voxels, which partition the 3D sensing region, and across multiple subcarriers. It exploits both spatial feature sparsity with channel diversity to pair voxel-level aggregation tasks and subcarriers to maximize the minimum receive signal-to-noise ratio among voxels. Optimally solving the resultant mixed-integer problem of Voxel-Carrier Pairing and Power Allocation (VoCa-PPA) is a focus of this work. The proposed approach hinges on derivations of optimal power allocation as a closed-form function of voxel-carrier pairing and a useful property of VoCa-PPA that allows dramatic solution space reduction. Both a low-complexity greedy algorithm and an optimal tree-search algorithm are then designed for VoCa-PPA. The latter is accelerated with a customised compact search tree, node pruning and agent ordering. Extensive simulations using real datasets demonstrate that Spatial AirFusion significantly reduces computation errors and improves sensing accuracy compared with conventional over-the-air computation without awareness of spatial sparsity.

Over-the-Air Fusion of Sparse Spatial Features for Integrated Sensing and Edge AI over Broadband Channels

TL;DR

A novel framework, called Spatial Over-the-Air Fusion (Spatial AirFusion), which exploits radio waveform superposition to aggregate spatially sparse features over the air and thereby enables simultaneous access and optimally solving the resultant mixed-integer problem of Voxel-Carrier Pairing and Power Allocation (VoCa-PPA).

Abstract

The 6G mobile networks feature two new usage scenarios -- distributed sensing and edge artificial intelligence (AI). Their natural integration, termed integrated sensing and edge AI (ISEA), promises to create a platform that enables intelligent environment perception for wide-ranging applications. A basic operation in ISEA is for a fusion center to acquire and fuse features of spatial sensing data distributed at many edge devices (known as agents), which is confronted by a communication bottleneck due to multiple access over hostile wireless channels. To address this issue, we propose a novel framework, called Spatial Over-the-Air Fusion (Spatial AirFusion), which exploits radio waveform superposition to aggregate spatially sparse features over the air and thereby enables simultaneous access. The framework supports simultaneous aggregation over multiple voxels, which partition the 3D sensing region, and across multiple subcarriers. It exploits both spatial feature sparsity with channel diversity to pair voxel-level aggregation tasks and subcarriers to maximize the minimum receive signal-to-noise ratio among voxels. Optimally solving the resultant mixed-integer problem of Voxel-Carrier Pairing and Power Allocation (VoCa-PPA) is a focus of this work. The proposed approach hinges on derivations of optimal power allocation as a closed-form function of voxel-carrier pairing and a useful property of VoCa-PPA that allows dramatic solution space reduction. Both a low-complexity greedy algorithm and an optimal tree-search algorithm are then designed for VoCa-PPA. The latter is accelerated with a customised compact search tree, node pruning and agent ordering. Extensive simulations using real datasets demonstrate that Spatial AirFusion significantly reduces computation errors and improves sensing accuracy compared with conventional over-the-air computation without awareness of spatial sparsity.
Paper Structure (33 sections, 5 theorems, 26 equations, 6 figures, 1 table, 2 algorithms)

This paper contains 33 sections, 5 theorems, 26 equations, 6 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. System Models
  3. Agent Perception Model
  4. Cooperative Sensing Model
  5. Communication Model
  6. Performance Metric
  7. Spatial AirFusion Protocol and Operations
  8. Sparsity Feedback
  9. Radio Resource Allocation
  10. Over-the-Air Fusion
  11. VoCa-PPA: Problem Formulation
  12. Greedy VoCa-PPA Algorithm
  13. Greedy VoCa Pairing
  14. Greedy VoCa-PPA
  15. Complexity Analysis
  16. ...and 18 more sections

Key Result

Lemma 1

Given the VoCa pairing matrix $\mathbf{A}$, setting an equal SNR level across all voxels, i.e., $\gamma_v = \gamma^{*}(\mathbf{A})$ for all $v$, is optimal for Problem P1, where $\gamma^{*}(\mathbf{A})$ is given as Substituting $\gamma^{*}(\mathbf{A})$ into eqn: pre-coding yields the optimal transmit power of each agent over a subcarrier,

Figures (6)

  • Figure 1: (a) An ISEA system for environment perception in the context of autonomous driving. (b) Spatial AirFusion protocol.
  • Figure 2: An example of a search tree for the optimal solution of Problem P2, with maximum depth, i.e., the number of agents, $K=3$. Nodes pruned by Proposition 2 are marked with strides.
  • Figure 3: The performance of variants of Spatial AirFusion and naive AirComp on the synthetic dataset.
  • Figure 4: The performance of variants of Spatial AirFusion and naive AirComp on the OPV2V dataset with number of CAVs $K=3$.
  • Figure 5: The performance of variants of Spatial AirFusion and naive AirComp on the OPV2V dataset with number of CAVs $K=4$.
  • ...and 1 more figures

Theorems & Definitions (6)

  • Lemma 1: Optimal Power Allocation
  • Lemma 2
  • Proposition 1: Stopping Condition
  • proof
  • Lemma 3
  • Proposition 2