Fire situational awareness wheeled robot system based on Federation learning
Jiayi Pan, Dongjie Wei · 2024
Fire situational awareness is essential for responding to fire events in a timely manner. However, the traditional fire detection system is often based on centralized data processing and decision making, which is easy to be affected by single point of failure and communication delay. In this paper, a wheeled robot system for fire situational awareness based on federated learning is proposed to improve the efficiency and reliability of fire detection and emergency response. The system consists of IoT systems in smart cities and wheeled robots that detect anomalies in the environment in real time through collaborative learning and autonomous movement of robots. In terms of robot architecture, we have adopted an innovative design scheme combining five-bar linkage legs structure and dual-wheel foot structure, so that the robot can walk freely in complex terrain, such as corridors and steps. The experiment shows that FSAR system has good performance in fire target detection, which is similar to the performance of centralized learning. Moreover, in the real world, the actual performance of the robot is consistent with the simulation results, and the robot can ensure the stability of the attitude under the condition of acceleration.