Like many agricultural markets around the world, the Australian farming sector is adopting semi and fully autonomous tractors along with Ag-Robots. The reason behind this adoption is similar to most markets, from labour shortages to the need to reduce input costs by adopting more precision farming practices. Innovative pricing models and the rise of “as a service” robotics has seen one of the traditional adoption blockers, that of cost, largely removed. This leaves one major blocker to the path of Ag-Robotic and autonomous machinery in Australia and that is the lack of connectivity.
Australia: a massive land country with a poor cellular coverage
With a very large land mass (Australia is just over 75% the land mass of the USA or about twice the land mass of Europe) combined with a low and mainly coastal living population the existing cellular networks while covering high percentages of the country’s population do not cover much of actual farmland. Around 65% of farmland in Australia has no or very poor cellular coverage and that coverage is a mix of 3G and 4G with only minor 5G to date. 2G has been switched off and the major national carrier, Telstra, is switching off 3G over the next two years so they can re-use the spectrum (frequencies) for 5G. This means in the interim and possible on going the 4G and 5G coverage is not equivalent to existing 3G levels making the connectivity problem over farms even more acute.
The connectivity problems fall into two classes – farms where there simply is no coverage and those farms where the coverage is patchy with areas of farms having no coverage at all and both situations cause problems for the deployment of ag robots and autonomy in farming. Clearly where there is no coverage the problem is easily identified however where there is patchy coverage it can be much harder to identify (See diagram 1).
The main effect of poor or no coverage often relates to the machine detecting an obstacle in its path and needs a “go/no go” commend from a remote operator – so the robot or autonomous tractor is detecting by LIDAR and / or its cameras that it can’t proceed as it detects an object which could be a person, other machinery, stock (animal), wildlife or even a large weed (yes they grow very quickly in Australia) – if the machine cannot communicate it will literally sit there until the object moves or the remote operator notes the machine is not where it should be and goes to find it. This sounds simple to solve with many radio technologies however the sheer scale of Australian broadacre cropping farms means this is not the case and in horticulture such as orchards, wireless signal penetration through the tree canopy can be an issue.
The issue is somewhat compounded by the regulatory settings in Australia – there are very low power limits set for class licenced spectrum devices such as UHF radio which means the effective range for data carriage is very low and placed based spectrum for use in private apparatus for 4G/5G networks is in the mid to high band zones meaning again range is somewhat compromised as is the ability of the signals to penetrate foliage and other local obstructions.Low band spectrum use is limited as it was allocated to the existing cellular operators in national lots, and they have little appetite to share the spectrum even in areas they do not use it. Australia does not have a spectrum use setting like the USA’s CBRS which limits options for private 4G/5G. Other technologies such as outdoor and mesh Wi-Fi (used a lot in Australian mining operations) tend to have deployment issues over farms due to the large areas needing covered and they have much smaller coverage areas.(See diagram 2).
On vehicle on the move satellite has shown some promise however it is a very expensive solution in use and has severe limitations in horticulture, especially in orchards where the tree canopy will likely block signals.
Australian agriculture operations using autonomy and robotics are often faced with a compromise of use efficiency where the machines may not be operating to their full capability and capacity due to the connectivity issue. To get round this several solutions are in play – some better than others. At a basic level the use of high gain antennas is a first step – effective in broadacre (wheat, barley cotton etc) operations however when we get into the horticulture environment there are many obstructions from tree branches to netting systems where antenna height on machines can be a problem. Flat form antennas are often used however their gains are somewhat lower than larger “whip” omnis. Stepping up from antenna choice is the use of high gain repeaters – the choice is limited to those that have both regulatory and cellular operator approvals – in the most this is Cellfi products. These are effective at amplifying the existing signal and they do that well – however they do need a signal – if there is no signal there is nothing to amplify, they are completely inert and ineffective. Boosters of any kind are not permitted in Australia by regulation due to the high levels of interference they cause in the cellular operators’ equipment settings.
The solutions continue in options with outdoor Wi-Fi over type land in its many variant forms of techniques which are mainly mesh. These do provide a workable solution however the very limited range of Wi-Fi often means that only partial farm coverage is achieved or there is a much higher number of deployed infrastructure elements involved.Private 4G and 5G networks do provide good connectivity for the operation of ag-robots and autonomous systems – with the added advantage that most of the routers in use, certainly in autonomous tractors and in most robots are compatible and are enabled by inserting the required SiM card. The issue for such private cellular networks is the availability of spectrum suitable for use over farms – however the 1800MHz and 2100MHz that is available as place-based apparatus licenses in many areas does produce reasonable results albeit at a higher cost than if low band 850/900MHz type spectrum bands were available as around double the number of towers and base stations are required.
Other aspects that have historically made the deployment of private networks in regional and remote Australia are power and network backhaul. Power has been resolved by the use of renewables with very effective solar, small wind and battery hybrid options and network backhaul – how the network “plugs in” to the rest of the world – has been resolved by the recent development of Low Earth Orbit (LEO) satellites such as Starlink’s Cellular Backhaul service.These LEO options offer very low latencies compared to other satellite option and if latency is not an issue for the robot or autonomous operating systems then there are other options such as the NBN business use satellite service.
In active use cases, Connected Farms has connected a range of differing robotics and autonomous tractors over varying types of Australian farm operations and have found that the use of private 4G (and in some cases 5G) is by far the most satisfactory option as it allows the machines to work at maximum efficiency. We have also seen the need, particularly with the robotics as a service model, to enable the means of connectivity to be portable – so that the connectivity can travel with the robots from farm to farm and have resolved that issue – this has taken a lot of focus as the previously mentioned place-based apparatus licences are geo-located however we have developed a very effective system for this.
Developments for future Australian farm connectivity to enable ag-robots and autonomous tractors will likely include more private networks. 5G will allow the use of much larger bandwidth however much of the 5G frequencies such as those in the 26/28GHz (mmWave) range have very limited range so their use in field robots will likely be more around mass data transfer / off load at or near farm hub or when near robot service vehicles (such as when refuelling or refilling spray tanks). This will allow robots to fulfil other duties as they carry out their primary role, for example a sprayer may also carry high resolution cameras and scanners to search for disease or pest infestations as well as crop quality – these functions requiring large bandwidth that would not be economical to deploy over a whole farm. Future developments in machine learning and AI may help with obstacle recognition however there are a few Australia specific problems to solve – to a machine a kangaroo is similar in shape and size to an adult human and the same applies for wallabies to children.
More effective layers of ubiquitous coverage over farms will also allow developments of taking some of the CPU needs off the robots and into localised edge computing facilities. This has the advantage of reducing the amount of expensive and sensitive computing equipment carried on-board and more effective use of that computing power with several machines being able to be serviced. This goes to help solve one of the problems of Ag-Robotics in Australia which is that of heat load. The robotics computing systems need to be kept in a clean environment and farms are very dusty environments during the many operation robots would be used for – which means that they tend to be is compartments on the machine that are pretty well sealed.Given the high ambient temperatures in a lot of areas of Australia combined with the heat load the onboard computing develops the environmental management can be challenging so removing some of that CPU to an edge facility with well controlled environmental systems would be advantageous.
In summary the adoption of ag robotics and autonomous tractors faces a connectivity issue like many other world markets – however the vast areas of the country that have no existing cellular coverage combined with spectrum regulatory settings mean that particular connectivity solutions are needed .The solutions that have evolved are being driven by specialist connectivity providers using a range of technologies generally either Wi-Fi or private 4G and these have been shown to work effectively though there has been some steep learning on the journey. The future use of 5G will see these specialised private networks supplemented in localised zones of very high capacity and evolutions where the coverage will itself become portable and move from farm to farm along with the robot fleets.It is hoped that future low band spectrum allocations may take account of the needs of agriculture and if the 600MHz spectrum band is better distributed around its current TV broadcaster use then dedicated spectrum could be made available. The agriculture sector in Australia is ready and willing to embrace autonomous implements and Ag-Robotics.The country has a growing supply domestic ag-robot and aftermarket autonomous tractor kits as well a few of international Ag -Robot companies starting to appear. The solutions for connectivity are not easy but with co-ordination between the robotic manufacturer, the farm and the connectivity provider an effective solution can be provided to enable maximum effective agricultural robot use.