Theory of mind (ToM) has been recognized as one of important cognitive functions for human beings and allows us to understand/model other's mind from their facial expression, verbal conversations, and behaviors. It has been known that the functionality has been developed in the early stage of our life. Recently, there have been some works on developing automated algorithms for robotic theory of mind. Similar to human's ToM, the process collects information on other robots and infers the internal states of them using estimation-exploration algorithms (EEA). Although they're successful, they assume that the playground has only one simple target to pursue. In real-world settings, the environment has several targets giving confusion to observers. In this study, we attempt to test the robotic theory of mind in the existence of multiple goals. Experimental results show that there are some successful conditions to get the best theory of mind capability for robots with a noisy target. This study gives insight on the human's theory of mind robust to the existence of multiple goals.