It is an inevitable industry trend to apply widely intelligent robots for molten steel sampling, because the steelmaking environment, characterized by extreme heat and airborne particulates, is hazardous to workers. However, it is a challenging problem to realize the autonomous path planning and stable control of robots. A path planning approach merging genetic algorithm (GA) and ant colony optimization (ACO) for molten steel sampling robots (MSSRs) is presented in this paper, which is named GA-ACO. Under the framework of GA, based on ACO, initial population generation, selection, crossover and mutation are improved. An intelligent PID controller with XGBoost optimizer is designed to guarantee the stability of MSSRs moving along the planned path. Simulation results show that generation average (GEN_AVR) and output average (OUT_AVR) of the GA-ACO are better than that of GA. The presented PID-XGBoost controller has higher control performance than the PID controller. The average error obtained by using the PID-XGBoost controller is -0.008m, and the average error obtained by using the PID controller is − 0.023m.