<p>In tight micro-assembly environments, particularly those with narrow passages and numerous interactions, the path planning of collaborative robots using the classic RRT* algorithm faces challenges such as long planning time and slow convergence. To address these issues, the IPQ-RRT*-Connect method is proposed to efficiently complete the 3D path planning of the tool center point (TCP) of collaborative robots in tight environments. First, the method integrates the greedy extension strategy from RRT*-Connect and the goal-biased strategy into PQ-RRT* to enhance the efficiency of obtaining a better initial path. Second, an adaptive oblique cylindrical sampling constraint scheme based on Informed-RRT* is proposed to facilitate rapid convergence in 3D tight space. In addition, a segmented constraint sampling strategy is introduced to enhance sampling performance for sharp turning scenarios caused by bypassing obstacles and narrow passages. Finally, the proposed method is compared with Improved P-RRT*, PQ-RRT*, and RRT*-Connect in Unity. The results show that the proposed method can search the best initial path efficiently, reduce the optimal path cost by 11% compared to Improved P-RRT* and RRT*-Connect, and converge to the <i>T</i>_5% path 67% faster than PQ-RRT* in tight micro-assembly environments.</p>