Effect of algorithm-based feed allocation on performance, health, and carcass outcomes of Brahman cross steers

• Feedlot bunk management can be automated with a lidar bunk scanner and feed allocation algorithm to achieve performance similar to a human managed process. • Automated bunk management relies on accurately determining pen feed consumption. • Automation of bunk management must be adaptive to changing environmental conditions and animal physiology. • Future iterations of automated bunk management algorithms could be improved by incorporating real-time weather data and refining steps to reduce pen-based daily fluctuations in feed intake. Successful feedlot bunk management (BM) is underpinned by human evaluations. Erroneous feed allocations can have profound impacts on performance and health across the feeding period. The objective of this study was to evaluate two levels of BM automation on performance, health, and carcass outcomes in a commercial feedlot. Bos indicus cross steers (n = 5509; initial BW = 420.2 ± 7.31 kg) were used in a randomized block design, consisting of three BM treatments comprised of 7 replicates per treatment. Treatments included: 1) semi-automated BM ( SEMI ), feed remaining were determined by the lidar bunk scanner and BM decisions were made by humans; 2) automated bunk management ( AUTO ), feed remaining were determined by the lidar bunk scanner and BM decisions were made via an algorithm; and 3) control ( CON ), where conventional BM decision were undertaken by humans. All treatments received the same feedlot induction health treatments including a hormonal growth promotant (HGP) implant containing 20 mg estradiol-17 β and 200 mg trenbolone acetate (Component TE200; Elanco Australia, Sydney), and were weighed and re-administered the same HGP implant between 42 and 45 days on feed (DOF). All steers were fed a steam flaked wheat and barley diet for 109 DOF, prior to slaughter at a commercial abattoir. There were no treatment effects for average daily gain (ADG), feed:gain ratio (F:G), final live weight, morbidity, mortality or hot standard carcass weight (HSCW; P>0.05). The CON treatment showed a tendency (P=0.06) for improved dry matter intake (DMI) over the total feeding period. The CON and SEMI treatments had greater DMI from DOF 1 to 33 (P=0.013). However, from DOF 34 to 66 and DOF 67 to 99, the CON and AUTO treatments had greater DMI than the SEMI treatment (P≤0.037). Outcomes from this study show that BM can be automated to achieve comparable performance, health, and carcass outcomes as highly-skilled human bunk callers. However, there were shortcomings identified with the automated system. The AUTO treatment had greater mean absolute deviation in DMI from DOF 34 to 66 and DOF 67 to 99 (P≤0.001), possibly due to larger and more frequent dry matter (DM) additions later in the feeding period. This suggests that human bunk management decisions were more adaptive as DOF increased and better accounted for changes in cattle physiology and DMI as cattle grew and matured. Moreover, this has highlighted a pathway for improvement in future iterations of the algorithm used.