711 Development of a Step Recognizing Device for an Autonomous Running Vehicle

Yutaka Tanaka, Akio Gofuku, Guangquan ZhangThe Proceedings of Conference of Chugoku-Shikoku Branch 2000 · DOI

Introduction

Autonomous mobile robots operating in human environments must reliably detect floor-level hazards such as stairs or ledges. While sophisticated solutions like ultrasonic arrays or stereo vision systems are available, they often demand significant computational resources and are costly. Researchers at Okayama University have addressed this by developing a high-speed, low-cost step detection system using simple infrared (IR) photo sensors coupled with an innovative automatic calibration circuit.

The Challenge of IR Sensing

The primary limitation of inexpensive IR sensors is their sensitivity to environmental variables. A sensor's output is influenced by the distance to the target, the angle of the surface, and the surface's reflectivity (color). A system calibrated for a light-colored floor may fail to detect hazards on a dark surface. This research focuses on an "Automatic Sensitivity Adjustment" mechanism that dynamically tunes sensor thresholds to the specific environment.

Methodology and System Design

The system utilizes Japan Autonics BA2M-DDT IR sensors. The core innovation is a custom circuit that manages binary logic signals (0 or 1) based on reflected light intensity.

Automatic Sensitivity Adjustment

To handle varying floor colors, the team designed a circuit incorporating an 8-bit hexadecimal counter and an analog switch network connected to a resistor ladder. Before operation, the system performs a "hunt" for the optimal threshold by stepping through resistance values until the sensor triggers for the current floor color. This value is stored in memory, effectively "zeroing" the sensor for that specific surface.

Geometric Optimization

Empirical testing was conducted to determine the optimal sensor geometry. The researchers found that a mounting height ($H$) of 340 mm and a projection angle ($\alpha$) of 60° maximized the detection range while maintaining high reliability across different surface types.

Figure 2 and 3: Effects of distance, angle, and color on sensor output, and the relationship between height, distance, and angle.

Key Insights

  • Analog Feedback Efficiency: By using a simple analog feedback loop to calibrate the sensor instead of complex digital image processing, the system significantly reduces latency and CPU overhead.
  • Inverse Detection Logic: Rather than searching for a specific "hit," the system monitors for deviations from a steady-state reflection. Any significant change in light intensity (increase for "up" steps, decrease for "down" steps) triggers an immediate response.
  • Strategic Placement: The 60-degree projection angle is a "sweet spot" that allows the sensor to distinguish between flat ground and a 3cm rise effectively, regardless of floor color.

Experimental Results

The system was validated using test environments with varying floor colors and physical obstacles.

  • Detection Capability: The device successfully identified "up" steps of at least 3 cm and "down" steps of at least 7 cm.
  • Calibration Precision: The auto-tuning circuit demonstrated exceptional stability, with resistance setting variances of only 0.2% for identical floor colors. The entire calibration process is completed in under 20 seconds.
Figure 5: Resistance values required for different floor colors (black, blue, white) for both up and down step detection.
  • Operational Performance: At a speed of 3 km/h, the robot achieved a full stop within 169–206 mm of detection. With a detection range of approximately 700 mm, this provides a substantial safety margin for avoiding hazards.

Conclusion and Future Outlook

This research demonstrates that high-performance hazard detection can be achieved with low-cost hardware through clever circuit design and geometric optimization. Future work may address challenges such as rapid transitions between non-uniform floor colors and the impact of intense ambient light on IR receivers.

Summary updated

BibTeX
@inproceedings{tanaka2000development,
  author = {Yutaka TANAKA and Akio Gofuku and Guangquan Zhang},
  title = {{711 Development of a Step Recognizing Device for an Autonomous Running Vehicle}},
  booktitle = {The Proceedings of Conference of Chugoku-Shikoku Branch},
  year = {2000},
  doi = {10.1299/jsmecs.005.1.239},
}
Figure