Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Safe mobility support system using crowd mapping and avoidance route planning using VLM

Sena Saito, Kenta Tabata, Renato Miyagusuku, Koichi Ozaki

TL;DR

This work addresses safe navigation of autonomous robots in crowded, dynamic environments by introducing an Abstraction Map Generator (AMG) that uses Vision-Language Models to detect abstract concepts like crowds and Gaussian Process Regression to convert these detections into a probabilistic crowd-density map. The crowd map is fused with a geometric map to produce a multi-layer cost surface, enabling Dijkstra-based path planning that avoids both static obstacles and crowds. Real-world campus experiments validate that AMG can generate usable cost maps and yield paths that circumvent crowds, while highlighting challenges in detection stability and the need for appropriate weight balancing. The study demonstrates a practical framework that bridges abstract scene understanding with geometric planning, with potential extensions to multi-layer environmental factors for enhanced robustness in real-time robotic navigation.

Abstract

Autonomous mobile robots offer promising solutions for labor shortages and increased operational efficiency. However, navigating safely and effectively in dynamic environments, particularly crowded areas, remains challenging. This paper proposes a novel framework that integrates Vision-Language Models (VLM) and Gaussian Process Regression (GPR) to generate dynamic crowd-density maps (``Abstraction Maps'') for autonomous robot navigation. Our approach utilizes VLM's capability to recognize abstract environmental concepts, such as crowd densities, and represents them probabilistically via GPR. Experimental results from real-world trials on a university campus demonstrated that robots successfully generated routes avoiding both static obstacles and dynamic crowds, enhancing navigation safety and adaptability.

Safe mobility support system using crowd mapping and avoidance route planning using VLM

TL;DR

This work addresses safe navigation of autonomous robots in crowded, dynamic environments by introducing an Abstraction Map Generator (AMG) that uses Vision-Language Models to detect abstract concepts like crowds and Gaussian Process Regression to convert these detections into a probabilistic crowd-density map. The crowd map is fused with a geometric map to produce a multi-layer cost surface, enabling Dijkstra-based path planning that avoids both static obstacles and crowds. Real-world campus experiments validate that AMG can generate usable cost maps and yield paths that circumvent crowds, while highlighting challenges in detection stability and the need for appropriate weight balancing. The study demonstrates a practical framework that bridges abstract scene understanding with geometric planning, with potential extensions to multi-layer environmental factors for enhanced robustness in real-time robotic navigation.

Abstract

Autonomous mobile robots offer promising solutions for labor shortages and increased operational efficiency. However, navigating safely and effectively in dynamic environments, particularly crowded areas, remains challenging. This paper proposes a novel framework that integrates Vision-Language Models (VLM) and Gaussian Process Regression (GPR) to generate dynamic crowd-density maps (``Abstraction Maps'') for autonomous robot navigation. Our approach utilizes VLM's capability to recognize abstract environmental concepts, such as crowd densities, and represents them probabilistically via GPR. Experimental results from real-world trials on a university campus demonstrated that robots successfully generated routes avoiding both static obstacles and dynamic crowds, enhancing navigation safety and adaptability.
Paper Structure (12 sections, 16 figures, 2 tables)

This paper contains 12 sections, 16 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Congestion Analysis Using Various Sensors
  4. Crowd-Based Navigation in Robotic Systems
  5. Abstraction map generation(AMG) method and path planning strategy with the map
  6. Confidence Evaluation of the Abstraction Map Using Gaussian Process Regression
  7. Environmental Maps
  8. Experiments on Crowd-Avoidance Navigation Using AMG
  9. Experimental Conditions
  10. Results of Map Generation Using AMG and the Resulting Planned Paths
  11. Discussion
  12. Conclusion

Figures (16)

  • Figure 1: Proposed our concept of Abstraction Map Generator(AMG)
  • Figure 2: Binary ocupancy of obstacle and crowd on the grid map
  • Figure 3: Smoothed gridmap by GPR
  • Figure 5: Robot used in the experiment
  • Figure 6: Geometric map in the experiment
  • ...and 11 more figures