The RRT algorithm is widely used in path planning due to its high efficiency in searching high-dimensional space. However, RRT will sample the entire workspace when sampling, resulting in many invalid sampling points, thus greatly increasing the planning time. To overcome these problems, we proposed OGRRT-Connect, an improved RRT algorithm based on opposite guided sampling, which has higher sampling efficiency and planning speed. Referred to the RRT-Connect algorithm, the algorithm expands two trees simultaneously. When the tree grows, the idea of triangular shrinkage guided sampling is introduced to limit the sampling area, improve sampling efficiency, and reduce planning time. Finally, we conducted robotic simulation experiments to prove the superiority of the algorithm in time planning and path nodes. Furthermore, we applied the algorithm to a real robot to demonstrate our proposed method.