Robots will change the way we do things, including the way we farm.

Robotic enabled weed control is an evolutionary step in precision agriculture and the dawn of weeding robots is upon us. Over the past 10 to 15 years a lot of investment in robotics research, and the tools that may be fitted to them, have occurred world-wide but few have been commercialised into relevant, practical and therefore functional farm assets. Lack of development investment and subsequent commercialisation are the current barriers to adoption of this technology.

Simply, an agricultural weeding robot consists of hardware and software - a self-steered (unmanned) and propelled platform that hosts an array of weed detection units that in turn activate an array of weeding tools whether that be a spray nozzle, microwave unit or tillage tool as examples.

Within Australia, four agricultural unmanned ground vehicles (UGV or robotic platforms) have been built. Most are still in the prototype testing stage; these include the Ladybird, RIPPA (both from Field Robotics Research Centre, University of Sydney), and AgBotII (Queensland University of Technology). Only one – Swarmbot (SwarmFarm Robotics) has been commercially launched and that occurred in late March 2016.

Currently there are only three weed detection/sensing units commercially available in Australia and they include Weedseeker and WeedIT, and more recently the H-Sensor originating from Europe. H-Sensor is able to detect green from green while the other two units detect green from brown which means the former is ideal for in-crop use and the latter are best utilised in fallow.

In it evident in the literature that many weeding tools and non-herbicidal tactics (eg. steam) have been developed but few have been commercialised or are not readily accessible for robotic application. Currently, microwave technology and a targeted tillage implement for weed control are being developed in Australia; the latter for the grains industry to tackle herbicide resistant weeds, but it too is not being developed for robotic enablement.

A brief overview of current weed management practices and the main weed issues of Australian cotton production systems have been described. Herbicide resistance development and species-shifts due to over-reliance on glyphosate arising from the advent of glyphosate tolerant varieties; as well as the management of rogue (volunteer and ratoon) cotton are the major current foci of cotton weed management. Based on this, the areas where robotic enabled weed management might fit have been identified. It is not the intent that robots should undertake all on-farm weed management tasks, but rather the more tedious and time consuming tasks that require maximum efficacy results. The key opportunities include but are not limited to:

• Spot spraying of weeds and rogue cotton plants in-crop and in fallow – low density situations or escapes of broadcast weed management applications; or odd patches of hard-to-kill and or resistant weeds; using alternate modes of action products.

• As above but by spot tillage (or other non-chemical treatment).

• Application via spot or patch spraying of effective but expensive herbicides that might not otherwise be used due to their cost.

• Weed surveillance – monitoring and mapping weeds in the paddock; the grower knows exactly where his weeds and issues are; monitoring for herbicide resistance survivors; becomes the eyes!

Finally, two specific areas where the CRDC might consider further investment in the short to medium term are suggested.