Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI-enhanced Direct SLAM: A Principled Approach to Unsupervised Learning in Bayesian Inference

Alexander Venus, Benjamin Deutschmann, Alexander Fuchs, Christian Knoll, Erik Leitinger

Abstract

In this paper, we propose an artificial intelligence (AI)-enhanced hybrid simultaneous localization and mapping (SLAM) method that performs Bayesian inference directly on raw radio-frequency (RF) signals while learning an environment model in an unsupervised manner. The approach combines a physically interpretable signal model for line-of-sight (LOS) components with an AI model that captures multipath component statistics. Building on this formulation, we develop a particle-based sumproduct algorithm (SPA) on a factor graph that jointly estimates the mobile terminal (MT) state, visibility, multipath parameters, and noise variances, and integrate it into a variational framework that maximizes the evidence lower bound (ELBO) to learn the neural network (NN) parametrization directly from measurements. We further present a highly efficient GPU-based implementation that enables parallel likelihood evaluation across particles and base stations (BSs). Simulation results in multipath environments demonstrate that the proposed method learns the generative, environment-dependent signal model in an unsupervised manner while accurately localizing the MT and effectively exploiting the learned map in obstructed-line-of-sight (OLOS) scenarios.

AI-enhanced Direct SLAM: A Principled Approach to Unsupervised Learning in Bayesian Inference

Abstract

In this paper, we propose an artificial intelligence (AI)-enhanced hybrid simultaneous localization and mapping (SLAM) method that performs Bayesian inference directly on raw radio-frequency (RF) signals while learning an environment model in an unsupervised manner. The approach combines a physically interpretable signal model for line-of-sight (LOS) components with an AI model that captures multipath component statistics. Building on this formulation, we develop a particle-based sumproduct algorithm (SPA) on a factor graph that jointly estimates the mobile terminal (MT) state, visibility, multipath parameters, and noise variances, and integrate it into a variational framework that maximizes the evidence lower bound (ELBO) to learn the neural network (NN) parametrization directly from measurements. We further present a highly efficient GPU-based implementation that enables parallel likelihood evaluation across particles and base stations (BSs). Simulation results in multipath environments demonstrate that the proposed method learns the generative, environment-dependent signal model in an unsupervised manner while accurately localizing the MT and effectively exploiting the learned map in obstructed-line-of-sight (OLOS) scenarios.
Paper Structure (24 sections, 36 equations, 3 figures)

This paper contains 24 sections, 36 equations, 3 figures.

Table of Contents

  1. Introduction
  2. System Model and Problem Formulation
  3. ai-augmented Measurement Model and State Vectors
  4. State-Transition Models and Prior PDFs
  5. Declaration and Estimation
  6. spa
  7. Prediction Messages
  8. Measurement Update Messages
  9. Particle-Based Implementation
  10. Efficient Calculation of Likelihood Function
  11. Variational Unsupervised Learning
  12. Inference Step (spa Algorithm)
  13. Learning Step (Unsupervised ai Model Learning)
  14. Data-driven Models
  15. nn Inputs
  16. ...and 9 more sections

Figures (3)

  • Figure 1: Concept figure of two time steps of the proposed method, where nn learn the environment to enhance the statistical measurement model. Top: factor graph and model-based components (measurement model in green) of the spa. Bottom: ai model in red (nn and other learned parameters) and their links to the signal and likelihood model. Quantities associated with unsupervised learning are shown in orange.
  • Figure 2: Graphical representation of the investigated synthetic experiment: Fig. (a) shows the simulated trajectory, the (single) known bs position, unknown walls (and respective va) and indices the olos situation.
  • Figure 3: Performance in terms of the rmse of the estimated mt position over time $n$ (a) and as the cumulative frequency of the magnitude error of the estimated mt position (b). The gray area represents the area of olos between bs and mt according to Fig. \ref{['fig:scenario']}.