Developing the use of ground robotics for data gathering and analysis to assist farming decisions
|Project start date:||01 June 2017|
|Project end date:||31 August 2018|
|Publication date:||01 December 2018|
|Livestock species:||Sheep, Goat, Lamb, Grassfed cattle, Grainfed cattle|
Download Report (1.5 MB)
There is growing interest in the need to provide timely information on pasture and animal state to livestock producers to support on-farm decision-making. The primary objective of this project was to demonstrate the application of mobile ground robotics for collecting high spatial and spectral resolution data of pasture and animal state on-farm. To achieve this, SwagBot, a prototype robotic vehicle developed by the Australian Centre for Field Robotics, was adapted and developed.
Extensive industry consultation was carried out prior and during the project to identify key areas within the production cycle where intelligent mobile robotics could support farmer decision making. These areas were identified as weed management, pasture survey and livestock monitoring.
An approach to autonomous driving on farm was demonstrated, making use of digital farm mapping, mission planning software and robotic perception technology for real-time obstacle detection and dynamic routing. An end-to-end solution for autonomous weed detection and spraying was demonstrated. The technical approach includes a computer vision system for real-time weed detection; a visual-servo control system to manoeuvre the vehicle into position; and a robotic spot spay system for chemical application.
Proof of concept pasture survey and livestock tracking applications were also demonstrated. In addition, the project showed interoperability with other emerging farm technology: data collected by SwagBot was used in combination with aerial data from a drone, and the mission planning system integrated with FarmMap4D, a commercial farm mapping tool.
Field trials on commercial livestock farms were used to validate the technology and to collect data for development of software algorithms. The locations of field trials were Chatswood, Nevertire and Arthursleigh, Marulan, both in NSW. The team also visited Central Queensland University’s livestock research facility to collect a dataset for development of livestock tracking software.
An assessment of technology readiness levels leads to the recommendation that the auto weed spraying technology is ready for commercial development now, with automated pasture surveying soon after. Livestock herding/leading was successfully demonstrated and a small amount of applied research will validate the technology. Longer term research would focus on livestock monitoring for welfare measurements.
|Project manager:||Nick Sangster|
|Primary researcher:||University of Sydney|