607 Development of autonomous omni-directional motor-driven cart for traveling in factory environment

Takashi Kuma, Shinichi NakajimaThe Proceedings of Conference of Hokuriku-Shinetsu Branch 2006 · DOI

Abstract

Researchers at the Niigata Institute of Technology have developed an autonomous, omni-directional transport cart designed to navigate cramped factory floors without the need for expensive guide rails or floor markings. By combining four-wheel omni-directional movement with ultrasonic sensors and a simplified avoidance algorithm, the cart can detect obstacles and maneuver around them while maintaining its course. This study demonstrates a flexible, low-infrastructure solution for dynamic industrial environments.

Introduction

In traditional factory environments, Automated Guided Vehicles (AGVs) often rely on "guide-ways"—physical tracks, magnetic tape, or painted lines—to navigate. These systems are rigid, expensive to install, and struggle in narrow, dynamic spaces where obstacles frequently block the path. To address these limitations, this research presents a more flexible solution: a cart that can move in any direction (forward, backward, sideways, or rotating on the spot) and autonomously navigate around obstacles using on-board sensors, eliminating the need for specialized infrastructure.

System Architecture and Methods

The system is built on three main pillars: hardware architecture, localization, and a custom avoidance algorithm.

Hardware and Control Flow

The cart's control system is centered around a PC interface that manages data acquisition and motor driving. As shown in the system block diagram, the PC communicates with an AD12-8(PM) board to interface with various components.

System architecture of the autonomous cart, showing the integration of sensors and motor drivers via the PC interface.
  • Orientation: Two 8-direction magnetic compasses (GEOSENSORY RDCM-802) are combined to provide 16-direction absolute orientation.
  • Obstacle Detection: Ultrasonic sensors (Best Technology US Sensor 2) are mounted on the corners to detect objects within a range of 0.5m to 3m.

Omni-directional Drive Kinematics

The cart utilizes four "omni-wheels" positioned at the corners. This configuration allows for holonomic movement, meaning the cart can translate in any direction without changing its orientation. The researchers developed kinematic equations to translate desired velocities ($V_x, V_y$) and angular velocity ($\omega_z$) into specific rotation speeds for each of the four motors.

Kinematic model of the four-wheel omni-directional drive system, defining the relationship between wheel speeds and cart velocity.

Localization and Avoidance

Instead of complex SLAM (Simultaneous Localization and Mapping), the cart estimates its position by integrating its velocity commands over time (dead reckoning). To minimize errors caused by wheel slip, a trapezoidal velocity profile was implemented to ensure gradual acceleration and deceleration.

When an obstacle is detected, the cart follows a scripted 8-step sequence:

  1. Advance until the obstacle is close.
  2. Correct position based on sensor data.
  3. Move sideways until the obstacle is cleared.
  4. Move forward past the obstacle.
  5. Return to the original path.

Key Insights

  • The Power of Lateral Movement: Omni-directional movement significantly simplifies obstacle avoidance. Unlike car-like robots with fixed turning radii, this cart can simply "sidestep" an obstacle, making the avoidance logic much more efficient in narrow corridors.
  • Managing Slip with Software: Wheel slip is a notorious problem for omni-directional wheels. The authors found that using a trapezoidal drive command significantly improved the accuracy of their dead-reckoning localization by avoiding jerky starts and stops.
  • Heuristic Avoidance: Rather than calculating complex potential fields in real-time, the authors used a robust, state-machine-based 8-step process to navigate around box-shaped obstacles, which is highly practical for structured factory environments.

Experimental Results

The researchers evaluated the system through direct physical experimentation:

  • Localization Accuracy: Over a 3m target distance, the error was roughly 2.3cm to 4.7cm. At a 10m target, the error increased to approximately 20cm in the direction of travel and up to 50cm laterally due to accumulated integration errors.
  • Sensor Precision: The ultrasonic sensors achieved a distance detection accuracy of 5–10cm within their operational range.
  • Functional Success: The 8-step avoidance logic successfully allowed the cart to detect a box-shaped obstacle, maneuver around it sideways, and return to its original goal path.

Limitations and Future Work

While successful, the current system relies on dead reckoning, which leads to accumulated error over long distances. Future iterations may require periodic "checkpoints" to reset position data. Additionally, while the current avoidance algorithm is optimized for box-like obstacles, future work will address more complex or moving obstacles and integrate higher-fidelity sensors like LiDAR to improve reliability in diverse factory settings.

Summary updated

BibTeX
@inproceedings{kuma2006development,
  author = {Takashi Kuma and Shinichi Nakajima},
  title = {{607 Development of autonomous omni-directional motor-driven cart for traveling in factory environment}},
  booktitle = {The Proceedings of Conference of Hokuriku-Shinetsu Branch},
  year = {2006},
  doi = {10.1299/jsmehs.2006.43.181},
}
Figure