Mobile Robot Indoor Positioning Based on a Combination of Visual and Inertial Sensors

Area of Science:

• Robotics
• Sensor Fusion
• Computer Vision

Background:

• Single-sensor navigation in mobile robots suffers from low accuracy and error accumulation.
• Multi-sensor integration is crucial for robust indoor navigation and positioning.
• Existing methods often struggle with limitations like sensor speed and precision.

Purpose of the Study:

• To develop an improved indoor mobile robot positioning method using a combination of visual and inertial sensors.
• To address the challenges of accuracy and error accumulation in single-sensor navigation systems.
• To enhance the reliability and precision of robot localization in indoor environments.

Main Methods:

• Utilized a Kinect sensor for visual data (color and depth images) and an Inertial Measurement Unit (IMU) for inertial data.
• Applied an improved Scale-Invariant Feature Transform (SIFT) algorithm for feature matching.
• Employed the absolute orientation algorithm to compute rotation and translation between image frames.
• Implemented an adaptive fade-out extended Kalman filter for loosely coupling visual and inertial data, optimizing position and attitude estimation.
• Collected real-time 3D data including acceleration, angular velocity, magnetic field strength, and temperature from the IMU.

Main Results:

• The combined visual-inertial sensor approach demonstrated significantly improved accuracy in indoor mobile robot positioning.
• The adaptive Kalman filter effectively compensated for sensor limitations, reducing cumulative errors.
• Experimental results validated the enhanced performance compared to single-sensor methods.
• Real-time data acquisition from both sensors provided comprehensive navigation information.

Conclusions:

• The fusion of visual (Kinect) and inertial (IMU) sensor data offers a superior solution for indoor mobile robot positioning.
• The proposed method effectively mitigates accuracy issues and error accumulation inherent in single-sensor systems.
• This integrated approach provides a robust and precise navigation solution for mobile robots in complex indoor settings.