Point Cloud Registration Algorithm Based on Improved Teaching-Learning-Based Optimization with Physics-Inspired Objective Function

Point cloud registration remains challenging due to efficiency, robustness, and global convergence trade-offs. Recent intelligent algorithms still face limitations in balancing efficiency, robustness, and global convergence. This study proposes an enhanced Teaching-Learning-Based Optimization (TLBO) method for 3D registration, incorporating a physics-inspired objective function. The algorithm achieves stable computation within 4-5 seconds while optimizing the efficiency-accuracy-versatility equilibrium. By resolving the conflicting demands between global robustness and computational speed, our approach maintains high-precision alignment across diverse scenarios. Experiments validate its effectiveness in operational stability and performance balance compared to existing methods, demonstrating significant improvements in handling complex registration tasks without compromising computational efficiency.