Prototype feed truck auto delivery refinements

Summary

A precedent project (MLA Project P.PSH.1079) delivered a world first auto-delivery system retrofittable to the mixer bins of commercial-grade delivery trucks, and this system was successfully prototyped at Bindaree Beef Group's (BBG) Myola Feedlot.

Through an agreed experimental protocol under normal operating conditions, the prototype and human operators were evaluated against several criteria for delivering steam-flaked wheat finisher ration. The precedent project's experiment outputs objectively demonstrate that the prototype system provided more accurate and more time-effective first pass and final deliveries than humans.

Key design features of the prototype system include its high-accuracy positioning solution based on RTK global navigation satellite system (GNSS) technology, it is retrofittable to feed trucks and scale heads that are commercially available, it can be switched between automatic and manual feeding modes, and the feed gate has an interlock preventing feeding errors if the feed truck arrives at the incorrect pen. While other commercial passive (operator assist) systems may already be available, we understand this is the first time an active computer-controlled system has been delivered and validated.

With a view to refining the working prototype as a more robust and commercially viable outcome, several improvements were required for the prototype system at Myola Feedlot. Most significantly, algorithmic refinements were required to enable more robust and reliable feed deliveries under normal operating conditions, and to include all rations, all masses, to all pen lengths, with minimal reversing.

After these control system refinements were completed, and appropriate operational confidence was garnered, an eight-week serialised experiment under normal operations evaluated the prototype system versus human operators. The review process considered performance criteria agreed to be relevant to all stakeholders. The distilled experiment results objectively show that the prototype system outperforms human operators with less reversing and therefore less distance travelled, and achieves higher accuracy in meeting total delivery allocation for target pen. The average travel speeds for each delivery mode were sufficiently different that direct comparison does not appropriately quantify time-based metrics, and it follows that extracting objective performance indicators is very difficult given the multi-variate nature of the problem. After the travel speeds for delivery sequences (i.e. travel within target pens) were normalised to the average human rate of

3.0kmh-1, it was determined that the prototype system enabled 19.2% faster feeding in tons per hour; for clarity, this is a proportional improvement of tons delivered per hour when the truck was in pens and outputting feed. The maximum benefit of this performance improvement was probably not fully visible, as the truck's utilisation (again, time in pen) was only 20.4% of its working day (observed, 16.5% after delivery speeds normalised), so the significant feed rate improvement is diluted by the majority of working time consumed by factors such as travel, reloading at mill and mixing delays, and external operating disruptions.

Based on the results of this project, the prototype auto-delivery system demonstrates exciting promise to improve feed delivery to bunks, with performances superior to human feed truck operators. In the short-term, our results suggest that even inexperienced or unskilled operators should now be able to deliver feed with high performances. The prototype system may also assist or complement future autonomous feed truck possibilities.