The Internet of Things, now generally known as IoT, promises to revolutionize agricultural decision-making. It’s exciting to think of deploying a vast network of sensors that can observe the on-farm environment in real time and help deliver best agronomic practices that can boost a grower’s bottom line, increase production sustainability and ensure freedom to operate. The reality of implementation can be challenging, however. Some issues may be expected, such as connectivity difficulties, but there are other hidden and progressively bigger obstacles lurking along the way. These challenges can be divided into different categories, each of which can grow to monster-sized proportions.
Different pieces of hardware and software have difficulty “talking” with each other, as manufacturers of the different components use different data formats, communications protocols, authentication schemes and so forth. This introduces friction and makes it hard to scale systems. For example, different loggers use different data formats, the cloud services that provide data from them use different names or codes for the same variables, and units of measure are rarely provided along with the data. This often all adds up to the data’s meaning getting lost in translation.
Working within established technology frameworks (e.g., AWS, IBM, Microsoft, Google, etc.) is helpful because the different parts or steps the data go through are laid out consistently within each framework, but system observability (i.e., understanding what’s going on in each part of the system) may be difficult or require proprietary tools. Some necessary features may not even be available, such as edge processing, where we perform calculations on the IoT device itself to reduce communications bandwidth or react to local situations. And operating costs may be difficult to estimate when starting out. Moreover, communicating data between providers’ proprietary clouds may be difficult.
We recommend implementing industry standard data exchange formats and using standard code lists whenever possible, as well as understanding the interoperability pros and cons of any IoT frameworks you consider adopting.
Malicious software and infrastructure attacks are becoming more frequent as well as more ingenious.
There is another major challenge beyond interoperability: security. The ag technology security landscape is far from ideal. At the device level, basic firmware, authentication/authorization and encryption are difficult to maintain, especially when corporate structure and resourcing are still catching up to IoT-specific requirements. When wireless technologies are used, we can expect greater risk of malicious attacks to company or grower assets, but field instruments typically transmit on limited-bandwidth connections, which makes the over-the-air firmware updates (that keep devices secure) expensive and difficult.
Then, once the data are collected, can we trust it’s really coming from our devices? Malicious software and infrastructure attacks are becoming more frequent as well as more ingenious. Many recent attacks on corporate infrastructure have exploited vulnerabilities in devices connected to the corporate network, such as digital cameras, DVRs, medical devices, etc. Devices considered for IoT should be accessible (remotely) for firmware, certificate and password updates. It’s also important to follow protocols for strong password, secure network and ecosystem interfaces.
Being able to securely move data from a device into the cloud while preserving its meaning and being confident of its origin isn’t enough. There is always the possibility of having “bad data” due to a breakdown in quality somewhere in the system. Old, damaged, dirty sensors; communications errors due to faulty antennae or overgrown line-of-sight transmission paths; and improperly identified sensors due to outdated configuration data on the logger or in the cloud are only some of the possible sources of data quality issues.
There is always the possibility of having “bad data” due to a breakdown in quality somewhere in the system.
The solution has two parts: The first is to define and continuously calculate and monitor data quality metrics that can provide early warning of data quality problems. For example, consider an indicator that shows a rain gauge that is consistently not showing rainfall despite rainfall appearing on nearby stations or gridded datasets. The second part is to put in place an aggressive preventive maintenance and inspection protocol that can target issues before they arise.
There is a huge fundamental difference between IoT as a hobby and developing IoT products for mass production. Although early-stage concepts in commercial applications are typically tested on hobbyist platforms such as Raspberry Pi or Arduino (for ease of development and the relatively low cost of obtaining a working prototype), these boards usually consume a lot of power and do not stand up well to challenging environments. The next step up is a ruggedized industrial processor-based development kit and, finally, custom-built circuit boards. Costs quickly escalate along this path.
Other scalability challenges arise from the chosen IoT data transmission technology. Topographic and land-cover variability on the farm limit connectivity and conflict with corporations’ requirements of low latency, independent and reliable transmission services for IoT. The throughput of different technologies varies widely, which can sometimes bring back interoperability issues. For example, some technology/frequency bands allow for barely 15 bytes of payload in unidirectional mode while others may allow 100 times faster, bidirectional throughput.
Managing the growth of IoT programs within the enterprise is key. Leadership will likely want to test the value that IoT can deliver, within a fail-fast context, but estimating the total cost of ownership of a system is very difficult before processes for estimating and managing the obstacles are put in place. Estimating the costs of preventive maintenance, dealing with faulty or compromised hardware, diagnosing and remediating communications problems, and gauging the system’s ability to satisfy the equipment’s power requirements in the field are good examples of early uncertainty.
Managing the growth of IoT programs within the enterprise is key.
One final aspect affecting scalability is communication and expectation management within the enterprise. The challenges posed and costs incurred by interoperability, security, data quality and scalability concerns are often not well understood — and not budgeted for — by leadership or by technical staff that may be asked to add IoT deployment and support to their existing responsibilities. Setting up robust internal IoT education programs, as well as supporting these programs from industry associations, seem critically important.
The “monsters” presented are not meant to keep users away from on-farm IoT. Instead, we hope to draw attention to them to help adjust expectations, avoid surprises and tackle scaling solutions based on the current state of the art. While undoubtedly there is active research in both academia and the private sector on how to make IoT technology more robust, user-friendly and easier to maintain and scale, a lot can be done with what is available today.
That being said, not all technology is created equal. Select technology that is based on standards, is built with interoperability in mind and follows industry standard best practices regarding security. We suggest putting processes in place to perform automated data quality assurance and understanding the maintenance logistics and power requirements (and your ability to deliver power throughout the season) of the devices you have in the field. Following these recommendations will go a long way toward successful, scalable IoT deployment.
8280 Willow Oaks Corporate Drive | Suite 630 | Fairfax, VA 22031
Tel: 703.536.7080 | Fax: 703.536.7019
HOME | ABOUT US | ADVERTISE | SUBSCRIBE | CONTACT | PRIVACY POLICY | IA ANTITRUST STATEMENT
