Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Are You Being Tracked? Discover the Power of Zero-Shot Trajectory Tracing with LLMs!

Huanqi Yang, Sijie Ji, Rucheng Wu, Weitao Xu

TL;DR

The paper tackles zero-shot trajectory tracing from unprocessed IMU data in AIoT. It introduces LLMTrack, a minimalist, role-play and Chain-of-Thought prompting strategy that enables LLMs to infer trajectories without domain-specific training. Empirical results on indoor and outdoor datasets show LLMs, particularly GPT-4 with CoT prompts, achieving an unseen average F1 around $80.1\%$, surpassing traditional ML and several DL baselines and demonstrating robustness to novel data. These findings suggest a promising, training-free path for integrating LLMs into CPS/AIoT systems for raw-sensor interpretation and trajectory analysis.

Abstract

There is a burgeoning discussion around the capabilities of Large Language Models (LLMs) in acting as fundamental components that can be seamlessly incorporated into Artificial Intelligence of Things (AIoT) to interpret complex trajectories. This study introduces LLMTrack, a model that illustrates how LLMs can be leveraged for Zero-Shot Trajectory Recognition by employing a novel single-prompt technique that combines role-play and think step-by-step methodologies with unprocessed Inertial Measurement Unit (IMU) data. We evaluate the model using real-world datasets designed to challenge it with distinct trajectories characterized by indoor and outdoor scenarios. In both test scenarios, LLMTrack not only meets but exceeds the performance benchmarks set by traditional machine learning approaches and even contemporary state-of-the-art deep learning models, all without the requirement of training on specialized datasets. The results of our research suggest that, with strategically designed prompts, LLMs can tap into their extensive knowledge base and are well-equipped to analyze raw sensor data with remarkable effectiveness.

Are You Being Tracked? Discover the Power of Zero-Shot Trajectory Tracing with LLMs!

TL;DR

The paper tackles zero-shot trajectory tracing from unprocessed IMU data in AIoT. It introduces LLMTrack, a minimalist, role-play and Chain-of-Thought prompting strategy that enables LLMs to infer trajectories without domain-specific training. Empirical results on indoor and outdoor datasets show LLMs, particularly GPT-4 with CoT prompts, achieving an unseen average F1 around , surpassing traditional ML and several DL baselines and demonstrating robustness to novel data. These findings suggest a promising, training-free path for integrating LLMs into CPS/AIoT systems for raw-sensor interpretation and trajectory analysis.

Abstract

There is a burgeoning discussion around the capabilities of Large Language Models (LLMs) in acting as fundamental components that can be seamlessly incorporated into Artificial Intelligence of Things (AIoT) to interpret complex trajectories. This study introduces LLMTrack, a model that illustrates how LLMs can be leveraged for Zero-Shot Trajectory Recognition by employing a novel single-prompt technique that combines role-play and think step-by-step methodologies with unprocessed Inertial Measurement Unit (IMU) data. We evaluate the model using real-world datasets designed to challenge it with distinct trajectories characterized by indoor and outdoor scenarios. In both test scenarios, LLMTrack not only meets but exceeds the performance benchmarks set by traditional machine learning approaches and even contemporary state-of-the-art deep learning models, all without the requirement of training on specialized datasets. The results of our research suggest that, with strategically designed prompts, LLMs can tap into their extensive knowledge base and are well-equipped to analyze raw sensor data with remarkable effectiveness.
Paper Structure (9 sections, 6 figures, 1 table)

This paper contains 9 sections, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. LLMTrack: zero-shot tracking with LLMs
  3. Experiments
  4. Dataset Setup
  5. Baseline
  6. Prompt Structure
  7. Evaluation
  8. Discussion
  9. Future Work and Conclusions

Figures (6)

  • Figure 1: Workflow of LLMTrack.
  • Figure 2: Experimental settings. a) Experiment devices. b)--i) Eight different scenes in indoor and outdoor environments.
  • Figure 3: IMU data visualization of two scenes including indoor turn right and outdoor turn right.
  • Figure 4: Chain-of-thought prompt design for LLMTrack.
  • Figure 5: Detailed step-by-step inference generated by GPT4 with a turn-left example.
  • ...and 1 more figures