In this paper we proposed a novel algorithm ant-based robot path planning (ARPP) based on ant colony system (ACS) to mimic a swarm of ants to find the globally optimal path for autonomous mobile robots. Visibility graph was used as both the roadmap and construction graph in ARPP. Although the near-optimums were readily available by ARPP, it is hard to find the optimality; therefore we proposed the special ARPP (S-ARPP) algorithm as its supplement. Experimental results show S-ARPP outperforms ARPP for higher qualities of the global-best path it found, and a flexible trade-off between satisfactory solutions and the number of iterations was also easily available to meet the certain needs.