SwarmFarm Robotics, an Australian robotics company, has three primary objectives:
- Deliver a flower and canopy mapping system that can capture the variability within an orchard block
- Create a decision support tool to visualise and understand the variation and give growers the tools to make better chemical thinning management decisions, and
- Develop a variable rate sprayer that can automatically change application rates throughout a block as pre-defined by the grower based on canopy and flower variation within the orchard.
These goals are part of the ‘AP16005 – Developing Agri-Tech Solutions for the Australian Apple Industry’ for Horticulture Innovation Australia, with a number of other project partners, including UNSW, Bosch and ADAMA to deliver an intelligent chemical flower thinning system to Australian apple growers.
Figure 1. Annie Xu (UNSW), Valarie Mengying HU (UNSW), Dr Mark Whitty (UNSW), Angus Ross (SwarmFarm) Angus Hogan (SwarmFarm), Dr Sally Bound (UTas) and Stuart MacLennan (SwarmFarm) with the Polaris data collection
SwarmFarm Robotics have also committed to deliver an autonomous solution to the horticulture industry in the form of their SwarmBot robotic platform, which they are already delivering to customers in broadacre cropping, cotton, turf, and macadamia orchards. Although SwarmFarm’s core product is robotic platforms, each of the systems they are developing for this project will be compatible with both a standard frame orchard tractor and a SwarmFarm robot.
SwarmFarm and its project partners believe that by quantifying and understanding the variation of flower load and the onset of uneven or prolonged flowering throughout an orchard block, and by utilising this data alongside a variable rate sprayer, growers will be able to more confidently target and correctly time chemical thinning applications to optimise crop load, reduce hand thinning, and ultimately improve fruit yield and quality.
Figure 2. SwarmBot “Indigo” testing along rows
This project is set to develop systems that are technically capable of mapping and managing flower load variations and deliver a viable product to industry. To ensure that these technologies reach market in a product form that delivers true value to growers, SwarmFarm is working closely with a range of tree crop industries and commercial partners to make sure that the spray technology is applicable to other industries.
Figure 3. Variable rate sprayer applying ATS to a row of Gala
Since May this year, SwarmFarm have been working in the Goulburn Valley developing and testing onsite at Hall M J & Sons in Tatura. The SwarmFarm team have been monitoring a block of Pink Lady and Gala varieties to test the data collection and spraying systems with the objective of testing sensor hardware and further iterating on the machine learning systems developed by UNSW.
To test these systems and improve the image processing systems that will automatically identify flowers in raw camera images, the team have been manually collecting large data sets daily. Following data collection scans, individual quadrants marked with QR codes positioned through the block are regularly counted for each stage of growth development (green tip, half inch green, pink bud, balloon stage, etc). This extensive data set will allow development of a system which will accurately quantify the flower load variation within a block.
Figure 4. Checking coverage of variable rate application of ATS in test rows
SwarmFarm have been working closely with Silvan Australia and Raven Industries, a global electronic control system manufacturer, to design and build a variable rate sprayer capable of independently controlling application rates between the top and bottom half of the canopy. For this season, the team worked with Hall’s Orchard and agronomists Michelle Egan and David Morey from IK Caldwell in Shepparton to manually define prescription spray application maps for the spray test rows.
Figure 5. SwarmFarm working with Silvan Australia and Raven Industries
The next phase of development will be to strengthen the data collection hardware, moving from the current season’s testing system to a modular unit. The SwarmFarm team will continue to operate the data collection system and variable rate sprayer throughout the growing season to continue to test the system and to ensure that the development of the spray system is suitable for year-round use for other spray applications. SwarmFarm will also be working with ADAMA to develop a decision support tool that will integrate the vision systems with spray map generation tools to automatically develop prescription maps for next season.
In a new development, SwarmFarm has recently signed a Memorandum of Understanding with Australian agtech company Green Atlas, to work together in a collaboration to further develop and improve flower counting technology for Australian growers.
Over the past few months SwarmFarm have been presenting at a number of industry events, and in the coming months will be hosting demonstration days and be presenting at the APAL Grower R&D Update event on the 13 November at the MCG in Melbourne.
To keep up to date with these events, see SwarmFarm’s progress and be involved in the discussion of the project on Twitter (@SwarmFarm) or Facebook (SwarmFarm Robotics).