<p>Currently, robots operating in real-world environments, such as homes and stores, require knowledge of their surroundings to perform tasks. This knowledge is represented as an environmental model, which is used for task planning and execution. Constructing an environmental model requires extracting features from the environment that are crucial for the task at hand. Active perception, which refers to the robot’s ability to actively interact with its environment to gather useful information, has gained attention as a technique for enabling robots to autonomously extract features through interactions, reflecting their physical embodiment. To achieve active perception, a robot must perform exploratory behaviors that facilitate interactions with the environment. Traditionally, such behaviors have been manually designed based on prior knowledge of the environment and the robot’s physical characteristics, limiting adaptability to new conditions. To overcome this, research has focused on enabling robots to autonomously learn exploratory behaviors. However, previous studies have mainly addressed object recognition, with limited research on environmental feature extraction. This study proposes a method for learning autonomous exploration behavior of a mobile robot to enable environmental modeling. The proposed method utilizes reinforcement learning to automatically acquire the robot’s exploratory behavior for obtaining sensor information, which is then used in environmental modeling. The effectiveness of the proposed method was verified through a simulation experiment using a mobile robot. Analysis of the sensor information obtained from the learned exploration behavior confirmed the extraction of both local and global environmental features. The hierarchical structure of the environment was also successfully represented. Furthermore, it was verified that the learned exploration behavior is effective in environments different from the initial learning environment.</p>