A Novel Real-Time Reference Key Frame Scan Matching Method.

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Artificial Intelligence

Background:

• Unmanned aerial vehicles (UAVs) are crucial for indoor search and rescue, but navigation in unknown, dynamic environments is challenging.
• Simultaneous localization and mapping (SLAM) methods often suffer from accumulated errors and high processing times.
• Existing point-to-point scan matching techniques are susceptible to outlier association errors.

Purpose of the Study:

• To propose a novel, low-cost, real-time 2D scan matching algorithm for UAV navigation in unstructured indoor environments.
• To mitigate accumulated errors and reduce computational load compared to existing SLAM approaches.
• To enhance the reliability and efficiency of UAV operations in search and rescue missions.

Main Methods:

• Development of a hybrid scan matching technique, Reference Key Frame (RKF), combining feature-to-feature and point-to-point methods.
• Utilizing a key frame strategy inspired by video streaming to minimize error accumulation.
• Integration of the Iterative Closest Point (ICP) algorithm for featureless environments and switching to RKF upon linear feature detection.

Main Results:

• The RKF algorithm demonstrated promising navigational and mapping performance in both static and dynamic environments.
• Significantly reduced computational time compared to various existing scan matching algorithms.
• Effective error mitigation and robust performance in unstructured and dynamic settings.

Conclusions:

• The proposed RKF scan matching method offers an efficient and accurate solution for real-time UAV navigation.
• Its low computational cost and improved performance make it suitable for resource-constrained UAVs in search and rescue.
• The algorithm shows significant potential for real-world deployment in complex indoor environments.