PHYSCENE is a novel method for generating physically interactive 3D scenes tailored for embodied agents. It integrates physics-based guidance mechanisms into a conditional diffusion model to ensure realistic layouts, articulated objects, and rich physical interactivity. The method addresses the challenge of generating scenes that adhere to physical constraints while enabling object interactivity, which is crucial for embodied AI tasks. PHYSCENE uses a conditional diffusion model to learn scene layouts and incorporates physics-based guidance functions to enforce constraints such as object collision avoidance, room layout, and object reachability. These constraints are converted into guidance functions that can be integrated into the diffusion model to ensure the physical plausibility and interactivity of generated scenes. The method is evaluated against existing state-of-the-art scene synthesis methods, demonstrating superior performance in terms of physical plausibility and interactivity. PHYSCENE achieves state-of-the-art results on traditional scene synthesis metrics and significantly enhances the physical plausibility and interactivity of generated scenes compared to existing methods. The method is designed to generate scenes with realistic layouts and interactable objects, making it suitable for a wide range of embodied AI tasks. The results show that PHYSCENE can effectively generate scenes that are both visually realistic and physically plausible, paving the way for further advancements in embodied AI research.PHYSCENE is a novel method for generating physically interactive 3D scenes tailored for embodied agents. It integrates physics-based guidance mechanisms into a conditional diffusion model to ensure realistic layouts, articulated objects, and rich physical interactivity. The method addresses the challenge of generating scenes that adhere to physical constraints while enabling object interactivity, which is crucial for embodied AI tasks. PHYSCENE uses a conditional diffusion model to learn scene layouts and incorporates physics-based guidance functions to enforce constraints such as object collision avoidance, room layout, and object reachability. These constraints are converted into guidance functions that can be integrated into the diffusion model to ensure the physical plausibility and interactivity of generated scenes. The method is evaluated against existing state-of-the-art scene synthesis methods, demonstrating superior performance in terms of physical plausibility and interactivity. PHYSCENE achieves state-of-the-art results on traditional scene synthesis metrics and significantly enhances the physical plausibility and interactivity of generated scenes compared to existing methods. The method is designed to generate scenes with realistic layouts and interactable objects, making it suitable for a wide range of embodied AI tasks. The results show that PHYSCENE can effectively generate scenes that are both visually realistic and physically plausible, paving the way for further advancements in embodied AI research.