Design and Application of a Cloud Platform for Broiler Inspection Robots

To reduce the labor intensity of manual inspection and support centralized robot management in commercial caged broiler farming, this study designed and implemented a cloud platform for broiler inspection robots based on a layered IoT architecture. The platform supports robot task scheduling, remote control, real-time status monitoring, multimodal data acquisition, hierarchical cloud storage, and structured warning-event management. Dead chicken detection was selected as the primary field-validated anomaly-analysis task. Open-mouth behavior recognition, thermal anomaly analysis, and abnormal sound identification were integrated as auxiliary screening functions. A three-month field trial was conducted in a commercial four-tier caged broiler house. The platform achieved a robot command success rate of 99%, a real-time status refresh delay below 3 s, backend service availability of 99.95%, and an average robot offline frequency of 0.2 times per week. Field validation of dead chicken detection achieved a precision of 90.6%, recall of 90.4%, F1 score of 90.5%, and cage-level localization accuracy of 96%. These results indicate that the proposed platform can support routine robot inspection, warning-event review, carcass localization, and targeted farm management response in commercial caged broiler production.