Fire as a Service: Augmenting Robot Simulators with Thermally and Visually Accurate Fire Dynamics

Abstract

Most existing robot simulators prioritize rigid-body dynamics and photorealistic rendering, but largely neglect the thermally and optically complex phenomena that characterize real-world fire environments. For robots envisioned as future firefighters, this limitation hinders both reliable capability evaluation and the generation of representative training data prior to deployment in hazardous scenarios. To address these challenges, we introduce Fire as a Service (FaaS), a novel, asynchronous co-simulation framework that augments existing robot simulators with high-fidelity and computationally efficient fire simulations. Our pipeline enables robots to experience accurate, multi-species thermodynamic heat transfer and visually consistent volumetric smoke without disrupting high-frequency rigid-body control loops. We demonstrate that our framework can be integrated with diverse robot simulators to generate physically accurate fire behavior, benchmark thermal hazards encountered by robotic platforms, and collect realistic multimodal perceptual data. Crucially, its real-time performance supports human-in-the-loop teleoperation, enabling the successful training of reactive, multimodal policies via Behavioral Cloning. By adding fire dynamics to robot simulations, FaaS provides a scalable pathway toward safer, more reliable deployment of robots in fire scenarios.

Figures (9)

• Figure 2: FaaS superimposes scene-aware renderings of fire on renderings of established robot simulation frameworks. Besides Isaac Sim our framework supports simulators such as Gazebo (left) and MuJoCo (right).
• Figure 3: FaaS Overview: The system asynchronously couples a conventional robot simulator (red boxes) with an external fire simulator (orange boxes) via a non-blocking data bridge. Taking a visual camera as an example, at each timestep, the robot simulator streams camera pose and scene information to the fire simulator. For the camera (blue boxes) we establish ROS topics that we use as a means of communication between the simulators. The fire simulation rolls out thermodynamic fire dynamics and renders a viewpoint consistent with the robot's camera pose. It then generates an alpha-matted image of flames and smoke. This rendering is composited onto the robot’s RGB observations and returned, producing geometrically and temporally aligned sensor data. The loosely coupled architecture preserves high-frequency robot control while enabling physically accurate fire behavior and real-time visual augmentation.
• Figure 4: Two Isaac Sim scenes. Top: A two-vehicle accident scenario. A large unconfined buoyant plume demonstrates open-air advection and wind-driven flame behavior around complex vehicle geometry. The left image shows an overview of the scene while the right image shows the scene from the perspective of the robot. Bottom: An indoor fire scenario. A constrained fire plume burning behind a counter. The smoke is contained by the ceiling and starts to expand into the room. The left image shows an overview, the right image the view from the robot.
• Figure 5: A fire burns behind a wall with a window. To obtain this scene we combine a rendering of a fire (c) with an RGB frame (a) and its corresponding depth map (b), which leads to a scene-aware fire rendering (d).
• Figure 6: Multi-fire scenario in Isaac Sim. Three combustion sources of increasing intensity produce distinct plume heights and asymmetric smoke volumes. This environment is reused for the thermally-aware path planning experiment (Sec. \ref{['sec:path-planning']}). Left: Overview. Right: Onboard camera view.