Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Omobot: a low-cost mobile robot for autonomous search and fall detection

Shihab Uddin Ahamad, Masoud Ataei, Vijay Devabhaktuni, Vikas Dhiman

TL;DR

This work tackles aging-in-place safety by introducing Omobot, a low-cost omnidirectional mobile robot that autonomously surveys indoor environments and notifies responders upon detecting a fall. The approach reframes fall detection as an object-detection task using YOLOv8-Pose and mitigates robot-viewpoint limitations with a homography-based transformation that aligns robot imagery with a human viewpoint, yielding 6–12% accuracy gains. It also advances motion control through motor-system identification, improving trajectory tracking and autonomous charging. The system is built on a cost-constrained, open-source hardware/software stack (under $700) and demonstrates end-to-end capabilities from mapping and navigation to fall detection and alerting, highlighting practical implications for eldercare devices and assistive robotics.

Abstract

Detecting falls among the elderly and alerting their community responders can save countless lives. We design and develop a low-cost mobile robot that periodically searches the house for the person being monitored and sends an email to a set of designated responders if a fall is detected. In this project, we make three novel design decisions and contributions. First, our custom-designed low-cost robot has advanced features like omnidirectional wheels, the ability to run deep learning models, and autonomous wireless charging. Second, we improve the accuracy of fall detection for the YOLOv8-Pose-nano object detection network by 6% and YOLOv8-Pose-large by 12%. We do so by transforming the images captured from the robot viewpoint (camera height 0.15m from the ground) to a typical human viewpoint (1.5m above the ground) using a principally computed Homography matrix. This improves network accuracy because the training dataset MS-COCO on which YOLOv8-Pose is trained is captured from a human-height viewpoint. Lastly, we improve the robot controller by learning a model that predicts the robot velocity from the input signal to the motor controller.

Omobot: a low-cost mobile robot for autonomous search and fall detection

TL;DR

This work tackles aging-in-place safety by introducing Omobot, a low-cost omnidirectional mobile robot that autonomously surveys indoor environments and notifies responders upon detecting a fall. The approach reframes fall detection as an object-detection task using YOLOv8-Pose and mitigates robot-viewpoint limitations with a homography-based transformation that aligns robot imagery with a human viewpoint, yielding 6–12% accuracy gains. It also advances motion control through motor-system identification, improving trajectory tracking and autonomous charging. The system is built on a cost-constrained, open-source hardware/software stack (under $700) and demonstrates end-to-end capabilities from mapping and navigation to fall detection and alerting, highlighting practical implications for eldercare devices and assistive robotics.

Abstract

Detecting falls among the elderly and alerting their community responders can save countless lives. We design and develop a low-cost mobile robot that periodically searches the house for the person being monitored and sends an email to a set of designated responders if a fall is detected. In this project, we make three novel design decisions and contributions. First, our custom-designed low-cost robot has advanced features like omnidirectional wheels, the ability to run deep learning models, and autonomous wireless charging. Second, we improve the accuracy of fall detection for the YOLOv8-Pose-nano object detection network by 6% and YOLOv8-Pose-large by 12%. We do so by transforming the images captured from the robot viewpoint (camera height 0.15m from the ground) to a typical human viewpoint (1.5m above the ground) using a principally computed Homography matrix. This improves network accuracy because the training dataset MS-COCO on which YOLOv8-Pose is trained is captured from a human-height viewpoint. Lastly, we improve the robot controller by learning a model that predicts the robot velocity from the input signal to the motor controller.
Paper Structure (35 sections, 1 theorem, 4 equations, 9 figures, 2 tables)

This paper contains 35 sections, 1 theorem, 4 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Fall detection using sensors
  4. Low cost mobile robots
  5. Robot Design
  6. Mechanical design
  7. Electronics design
  8. Microcomputer
  9. PCB design
  10. Battery management system
  11. Motors
  12. Sensors and communication modules
  13. Software design
  14. Ros-based Software Framework
  15. Robot setup phase
  16. ...and 20 more sections

Key Result

Theorem 1

Let the plane ${\cal P}$, on which the points $\mathbf{x}_1$ lie, be described by a unit normal vector $\hat{\mathbf{n}} \in \mathbb{R}^3$ and the distance $h \in \mathbb{R}$ from the origin so that the plane is defined as ${\cal P} = \{ \mathbf{x} \in \mathbb{R}^3 \mid \hat{\mathbf{n}}^\top \mathbf

Figures (9)

  • Figure 1: System overview.
  • Figure 2: From left to right: 1) 3D model of the robot developed in Solidworks, 2) PCB Assembly with the major components, 3) the final assembly of the robot.
  • Figure 3: Communication among sensors and devices with ROS nodes running on Jetson Nano.
  • Figure 4: Left: Trajectories of the robot. Right: Deviation from the expected trajectory with and without using the parameters from motor system identification.
  • Figure 5: Mapping using LiDAR, Wheel Odometry, and ros-slam_toolbox. Left: robot's and odometry's trajectory. The green arrows show the trajectory of the robot's origin while creating the map, and the red arrows show the trajectory of the odometry's origin. Right: the deviation of the robot's and odometry's pose (translation in XY axis, orientation w.r.t Z axis) from the map's origin.
  • ...and 4 more figures

Theorems & Definitions (1)

  • Theorem 1