By Ulf Kerer and Greg Dryden (First seen in The Orchardist, March 2014)

Unprofitability of orchards through rising costs and stagnating produce prices have shifted the research focus around the world on how to reduce labour costs and lower operating costs. One trend is to increase mechanisation to reduce labour costs or lift productivity in the orchard.

Some of the technical components are not new but are innovative in the way that they have been combined for the first time for use in an orchard. Some innovative companies envision a fully mechanised orchard where the only task of humans will be putting apples on conveyor belts at harvest. This could mean in the future that the orchard is not only pruned mechanically, but tasks like mowing, spreading fertiliser or application of agrichemicals is done exclusively by orchard robots. During the growing season canopy development, soil moisture and fruit development are monitored by independently driving robots collecting data. After another robot has delivered an exact yield estimate bins are placed throughout the orchard accordingly and once they are filled by pickers working with harvesting machines, bins are collected and transported to the yard with a minimal amount of labour. Sounds too futuristic? Most of the tasks described above can already be undertaken with machinery that is available or in the final stages of development.

Self-driving systems
Still advanced, but comparatively simple are several research projects trying to create a system where a tractor is guided through the orchard autonomously. The aim is to shift the driver’s attention from trying to drive a straight line to being more focused on the performance of machinery or sprayer. The overall aim of these projects is certainly to replace the worker in the cabin altogether. The ‘steering by wire’ method is already well established where the driver is not actually sitting in the cabin but controls the tractor with a remote control. This system saves an entire worker if a platform on a trailer is attached to the tractor. Projects in Switzerland came to the conclusion that a system that is solely based on GPS technology could prove too unreliable in an orchard. GPS based tractor guidance systems are well established in open farming, but driving in an orchard requires more adaption to terrain, posts or rows that are not perfectly straight. Projects in Switzerland used laser and ultrasound sensors that react to trunks and canopy as a guide for the tractor. Trials in orchards have shown that maximum variation is 10cm when driving several times in the same row. The sensors proved to be absolutely reliable even when driving at speeds up to 7 km/h. In that scenario the time from recognising an object 4m in the distance to reaction is only 0.12 seconds. The steering mechanism was an electro motor that was that was installed at the steering wheel. After use that device can easily be switched to one side and does not interfere with steering anymore.

Fruit Robot ‘Caesar’
The fruit robot Caeser has been developed specifically for use in orchards and vineyards. It has just been introduced to the market and currently it is still in a phase where it needs to be established what the full potential could be.  This machine is able to drive independently in pre programmed lanes between trees or vines. It has a weight of 1,600kg, 2 wheels and is 3 meters long. The system accesses GPS data to calculate the current position in the orchard, it then uses a radio-transmitted correction signal. According to the manufacturer this enables absolute and reproducible lane accuracy of 2 to 3 cm. That means that after the initial setup there is no further requirement for a worker unless it has to be brought to the orchard from a storage facility.
The installed safety features stop the vehicle immediately in the event of an obstacle. The robot can spray, mow, spread fertilizer or transport bins depending on what modular addition is attached to the basic vehicle. Two manufacturing companies have already agreed to modify their equipment (sprayers and harvesting machines) so it can directly be installed on the robot. An interesting feature is certainly that the fruit robot is not bound by ‘normal’ working hours but can be used around the clock whenever it suits the grower or conditions for the designed task are right.

3D Mosaic
This interesting robot has been coordinated by the ATB Institute in Potsdam, Germany over the last years and has recently been presented in Germany. It is an autonomously driving platform that travels through the orchard and collects data about tree and fruit development. The vehicle is equipped with 3D-, Hyper-spectral-imaging and Thermography-cameras that measure tree growth and canopy development. Also soil depth profiles are recorded at the same time and combined with the other data to create a geo-coded map of the orchard.
Additionally, selected fruits in the orchard can be attached to micro sensors that monitor a range of information relating to fruit quality. While driving through the orchard, data from these fruit is also collected and processed so that the grower can draw conclusions about fruit quality. Additionally the grower can determine the optimal harvest period or determine the right timing for irrigation. The futuristic long term goal of this project is to divide the orchard into sub-plots (or even individual trees) and establish the right amount of water or fertiliser for these. This system has been tested successfully in fruit orchards in Turkey and Germany and can be expanded into other sectors like grapes or crops grown in greenhouses. The follow-on projects will now try to create a model that it robust enough to be used in a commercial orchard and that can be used in field without the help of scientists

Automated Yield Estimation
The Carnegie Mellon University in Pittsburgh developed an interesting robot that is capable of driving autonomously through the orchard and delivering a yield estimate. The support vehicle was also developed at the same university and is equipped with a positioning system. It travels at a preset speed through the orchard and uses two high resolution cameras that are mounted on an aluminium bar for image acquisition. The system works at night time with controlled artificial lighting to avoid the variance from natural light. A computer detects the apples on the trees and creates a yield estimate for the entire block. The researches were faced with several challenges: natural light, hidden fruit and multiple counting of the same fruit. The system was trialled on Red Delicious and Granny Smith blocks in commercial orchards in Washington State. The automated yield estimate was compared to a conventional yield estimate by orchard workers. The results for Red Delicious were fairly accurate and consistent, but un-thinned trees were significantly undercounted. So were the Granny Smith trees – most likely due to the green colour of the fruit and thus invisible for the cameras. However further tests have shown that the error was consistent and therefore the system could be re-calibrated to deliver a more accurate yield estimate. After the data collection the system is also able to produce a yield map and show yield variance within the orchard.

Sensor technology for orchard sprayers
The driver for research and development in this sector was an absolutely targeted spray application without any loss, the elimination of spray drift and the reduction of adverse effects on the environment.  Infrared sensors send light that is reflected by parts of the tree. These sensors analyse the canopy and transmit the data to micro processors; the software then calculates the optimum amount of spray for the target area. This system is not only designed for missing trees or missing branches, but also for lighter canopy at the bottom or the top of trees. Nozzles targeting those areas get then switched off temporarily by a valve. The speed for the tractor needs to be calculated exactly and is the part of the system to allow an exact calculation. The tractor driver can see in real-time on a screen the amount of spray that is saved. An additional advantage is that at the end of the row the sprayer starts and stops exactly at the last tree and does not need to be turned off manually.

Conclusion
Undoubtedly most of these products or system are very exciting and have the possibility to have an impact on orcharding in the future. Some of these technologies have just been introduced to the market or undergoing further trial. It is difficult to predict whether some of these products will ultimately make it to commercial production and to success in the marketplace. It will be interesting to see the developments in the near future and to analyse the possibilities for New Zealand orchards. The key to whether any of these recent developments will be widely used in orchards around the world will be a thorough cost-benefit analysis. In some cases the there still seems to be a lack of reliable and independent data to make an informed decision but nonetheless as this work is completed no doubt there will be steady increase in robotic activity on our orchards!

Acknowledgements
The author would like to thank the following people for supplying the Photos:
Stephen Nuske (Carnegie Mellon University)
Helene Foltan (Leibniz-Institut für Agrartechnik Potsdam-Bornim e.V.)