A new bio-inspired meta-heuristic algorithm, named Sled Dog Optimizer (SDO), is proposed in this paper. The inspiration for SDO is primarily drawn from the various behavioral patterns of sled dogs. It focuses on constructing mathematical models by simulating the process of dog sledding, training and retiring behaviors. The mutual constraints of multiple steps and the variation of several parameters make the exploitation and exploration capabilities of SDO well balanced. To test the ability of SDO, this paper sets up several different sets of experiments to be analyzed from a number of different aspects. First, SDO is qualitatively analyzed through several performance metrics, which include search history, search trajectory and population diversity. Second, SDO is compared with several novel and excellent metaheuristics at CEC 2017 as well as CEC 2020. The results show that SDO has extremely good performance in solving unconstrained optimization problems. SDO is then further applied to the CEC 2020 real-world problem test set to test its ability to solve real-world optimization problems with constraints. Finally, this paper extends and proposes a new path planning model to which SDO is applied. In these problems with constraints, SDO performs well and has a promising application. SDO demonstrates outstanding control of exploration and exploitation capabilities in all experiments. It can be said that SDO promotes the progress of meta-heuristic algorithms and provides another new powerful technology for complex optimization problems in the real world.