Precision Livestock Farming (PLF) systems utilizing rail-mounted robots offer promising solutions for environmental monitoring and mortality detection in commercial poultry production. However, challenges remain regarding sensor accuracy and computer vision reliability. This study examined how sensor height placement and camera occlusion influence data accuracy and detection efficacy of the SCOUT® rail-mounted robotic system deployed in a commercial broiler house (16.5 × 152 m, Bradley County, Tennessee, USA). Environmental data (temperature and relative humidity) collected at robot sensor height (1.7 m) significantly differed from measurements at bird-level height (0.4 m), with lower temperatures (-1.4°C) and higher relative humidity (+10.6%) recorded at bird-level. Mortality detection accuracy was assessed by comparing robot-reported mortalities with manual counts over two production cycles, revealing an overall detection rate of 31%. Staged mortality tests clarified the impact of occlusion, revealing significantly higher detection rates beneath the rail when mortalities were isolated within enclosures (57%) compared to outside (19%; p < 0.001). Detection rates were substantially lower near feed and water lines and house sidewalls due to obstruction from infrastructure and live birds, yet still significantly higher within enclosures (p = 0.042 and p = 0.022, respectively). Additionally, camera field of view limitations (approximate 1 m effective range) and poor performance in low-light conditions further reduced system reliability. This research identifies specific limitations in current PLF robotic designs and proposes technological refinements in sensor positioning and vision systems to improve monitoring accuracy and bird welfare management in commercial houses.


 

broiler,Precision livestock farming,robot,welfare,rail-mounted