Laser weeding arrives in Australia, how does it work?

Tripod Farmers, which specializes in growing various types of salad using conventional methods, has recently been faced with increasing weed pressure. This is the result of crop intensification and limited rotation, leading to increased weed resistance.

In addition, since the last 10 years, the company has found it increasingly difficult to recruit labor, and the costs associated with employed or contracted labor have risen years after years. Faced with these challenges with no obvious solution in the traditional market and looking to reduce production costs while ensuring more sustainable maintenance of its fields, Tripod Farmers discovered laser weeding technology.

Shortly afterwards, it is at FIRA 2023 in California that the company was definitively won over and will take the decision to buy the machine at the end of 2023. At the time of our visit in early January 2024, the machine had just been delivered to the farm and had been operational for just 5 days, having had to contend with a few days of bad weather that delayed its use.

_{Text continues below picture}

5 hours, 5 acres, 800,000 weeds killed

Franky Ruffo, the farm manager and third farmers generation, told us he was surprised at how quickly the machine was up and running and how easy it was to use. “The models aren’t perfect yet, that’s what we’re working on with the Carbon teams, but in just a few days we’ve already tested the machine on four different crops, each with its own specific configuration, and the results are impressive.”

After just 5 hours of operation, the results were already significant: 5 acres weeded, and 800,000 weeds eradicated. This represents an average of 160,000 weeds eliminated per hour and 1 acre per hour (0.4 ha/h). For a launch, Franky is very satisfied. Given that Carbon Robotics announce a machine capacity up to 300,000 weeds removed per hour and productivity of up to 3 acres per hour (1.2 ha/h), Franky is aware of the scope for improvement and is confident about the future.

_{Text continues below picture}

An expected ROI of 3 years

With Franky hoping for a return on investment (ROI) within 3 years, the laser weeding machine looks set to be a major asset for Tripod Farmers. After five days of intensive trials on a variety of crops at different stages of development, Franky was impressed by the machine’s adaptability and the performance it offered in such a short space of time. This initial phase of testing on a variety of crops gave Franky a more in-depth view of the total surface that the machine would be able to cover per year.

With a view to achieving a rapid return on his investment, Franky is aiming to reduce his weeding labor requirements by 10 people over the next two years. He even envisages a reduction to 20 people from the third year onwards, once the machine, the working methods and the organization have reached their maximum potential.

Franky explains that each employee represents an annual cost of between 60,000 and 70,000 Australian dollars, or even more if he must outsource the work. By reducing his teams by 10 people, he expects annual savings of up to AUD 650,000. This is how he expects to quickly make his machine profitable, and these figures do not consider the potential savings linked to the reduction in the use of herbicides, an optimization that he is also considering over the next few years.

_{Text continues below picture}

Need an expert on site for the launch

As is often the case when expanding into new markets, Carbon Robotics comes up against local peculiarities. Variations such as soil type, crop configurations, and even the type of weeds present in the plots require the machine models to be readapted to meet the specific needs of each customer.

During the launch phase, Carbon deploys an agricultural technology implementation expert to the site to accompany the user, train them and work closely together to fine-tune the machine to the expected level of efficiency and productivity. This phase can last several weeks, depending on the complexity and diversity of the configurations in which the machine must operate. Mark Zahnlecker, product specialist at Carbon Robotics, stresses that it’s not the getting to grips with the machine and training the operators that are the most time-consuming elements, but rather the process of adjusting each model of the machine to suit the type of crop, to achieve optimum results. The machine, which is controlled from a iPad tablet supplied by the manufacturer, can be mastered in just a few minutes.

_{Text continues below picture}

Detection and eradication of a weed

At the time of our visit, Franky and Mark were still working on the model dedicated to growing rocket lettuce for the baby leaf market. The two main challenges they were facing were a non-uniform, non-straight emergence, due to the heavy weather of recent weeks, and the detection and eradication of a weed very similar to the crop itself. The model already recognized this weed but was not yet as effective as it could be.

The final big challenge for Franky is the logistics of the machine. The 6-meter-wide machine is not easy to move from one farm to another.

48 hours to rework their AI model!

During our visit, one of the observations that particularly caught our attention was the speed with which Carbon Robotics manages to train and generate new artificial intelligence models that are reliable and accurate. Their process seems to be completed in between 48 and 72 hours at the most, a notable efficiency compared with other companies in the sector, which generally take between two weeks and a month to complete the same task.

When we asked Mark about the method used by Carbon Robotics, he was relatively evasive: “with a vision-based system we can quickly capture images to integrate into our deep learning algorithms which have the ability to process and create new models within that timeframe”.