It is essential for smart house researchers to have large datasets from actual environments. However, not all researchers have sufficient budgets to build test beds. These researchers need a simulator that can synthesize realistic sensory datasets. To solve this problem, we propose the 'UbiSim' simulator for activity recognition research. UbiSim provides a 3D graphical user interface to enable spatial perception using multiple sensors, including those that detect motion, pressure, vibration, temperature, and contact, along with RFID tags and receivers. The sensors are designed to verify collisions in a virtual space as a means to operate with minimal computational costs. Our proposed methods were tested in a virtual environment. The results show that the smart house simulator achieves real-time performance.