Why does it matter?
Weed management in traditional crops is a well established paradigm, especially due to the fact that crops are grown in regular patterns. Conversely, in grazing situations the occurrence of weeds is random – leading to weed management protocols which are heavily manual and are neither cost- or time- effective. By harnessing the outstanding developments in the past few years within the fields of robotics, remote sensing and artificial intelligence, this project aims to package these capabilities into a cohort of autonomous robotic vehicles to implement a commercially feasible weed management solution outside of cropping environments.
The project aims to develop a robot to autonomously find and identify individual noxious weeds, spraying or using an alternative method to eliminate them. It will produce a series of operational prototypes (“Kelpie”), based upon a commercially available agricultural chassis. These systems will autonomously navigate pasture or vineyards to economically control weeds where it is currently not possible, detecting and avoiding obstacles without human intervention. It will be able to classify individual noxious weeds, such as Serrated Tussock, in a variety of conditions, and avoid damage to beneficial grasses.
Additionally, Kelpie will record the position and size of the weeds, to produce a Weeds Map for planning and audit purposes, and a Feed Quality Map with the current farm stock carrying capacity. The weeds data collected during the project will be analysed and made publicly available as a research resource as a Reference Weeds Library.
There are five project outcomes:
- Kelpie weed control robot proven in trials
- Kelpie Management System (KMS) that will coordinate multiple Kelpies
- Commercialisation plan
- Educational program
- Farmer outreach program, including Reference Weeds Library, a database of photographs classified by weed type.
QCR will conduct research into vision detection algorithms with the aim of detecting and classifying vegetation (both common weeds and beneficial grasses) as an input into the weed control system. QCR will also conduct research into and development of vision-based obstacle detection and localisation algorithms within pasture-based farms and inter-row between vines in vineyards. These algorithms will infer the presence of obstacles through detection, localisation and classification as an aid to the obstacle avoidance path planner.
Real world impact
Benefits to society are trifold:
- The Kelpie platform doesn’t require a human to operate it, so it assists in mitigating a critical shortage of available farm workers to conduct this type of task, with the impact of weeds estimated to cost Australian livestock industries $2.1Bn per annum in control costs and lost production (MLA 2019). With an ageing workforce, transitioning the industry to the businesses of tomorrow is a crucial component of future proofing the Australia agricultural sector.
- Its selective action reduces the collateral damage to pasture that occurs with boom spraying, and minimises chemical usage, which has negative environmental implications in terms of chemical run-off and contribution toward herbicide resistance species.
- It has the potential to develop into a sovereign capability with export opportunities, harnessing Australian SME’s into the supply chain whilst also providing circumstances for regionally based organisations to provide weed-treatment-as-a-service. As the technology matures, further R&D can occur to diversify the capabilities of the platform to include (for example) fence inspection, stock monitoring and logistics.
Funding / Grants
- Commonwealth Grant CRCPIX000099 (Cooperative Research Centre Project); Agent Oriented Software (Lead Partner) (2020 - 2022)