IoT and AI projects in Azure Cloud: an example of a transport company
If you take an interest in new technologies, you will hear buzzwords like “IoT”, “AI” or “Machine Learning” being thrown around a lot. Everyone knows them, but not everyone can really say what they mean or how they can be used in IT projects. Do you need difficult, specialist knowledge to tap into the opportunities of artificial intelligence or can easily available cloud services be of help? In this article we’ll take a look what Azure Cloud have to offer.
What is IoT and how is it used in daily life?
Briefly put, the IoT (Internet of Things) is a network of interconnected devices, i.e. anything that can be fitted with a simple operating system that allows users to control its parameters. As the popularity of the IoT grows, we are beginning to see more and more smart fridges, smart coffee machines, smart light bulbs…and the list goes on.
When you have a set of devices, all you need to do is connect them in a local or global network. What sets them apart is their set of instructions. They may be fairly autonomous; they may collect certain types of data for future analysis, and even process and share them with one another.
While it may be difficult to understand why a smart light switch would need to know how many coffees you prepare per day, all the pieces begin to fall together when you have, e.g., two smart vacuum cleaners. If they are to clean up a single space, they need to divide their tasks somehow, and this is where data-sharing comes in handy.
Artificial Intelligence vs Machine Learning
In contrast, AI, or artificial intelligence, is a more intricate term, in part because we have not developed real artificial intelligence just yet (and it may still be some time before we do). What we do have at the moment is a way to create “learned” behaviour, a certain function based on a large quantity of input and output data.
For example, if we have a data set that contains:
- as an input: millions or billions of photos of text,
- as an output: text in these photos,
we can develop a machine-learning model that would work through the data set and then easily convert any text excerpt from photo to text format.
This is precisely what Google did through its Captcha project.
Cloud computing in IoT and AI projects
How do these two strands come together in the cloud? Azure Cloud and other providers, such as AWS or Google Cloud, create their own solutions based on these concepts and provide API or user interfaces to facilitate and speed up their development. Usually, these are ready-made products for e.g. object recognition in photos, speech recognition, or comprehensive IoT work environments. Of course, products of this kind come with many advantages offered by the cloud, such as availability, resources, and scalability.
But why should you use ready-made products at all? First, this will allow you to ensure the security of your cloud solution. Since the products are already in use in thousands of other companies, you know that they have already been thoroughly tested. The risk of vulnerability in this case is much lower than if we were to create everything from scratch. In addition, developing similar functionalities is very time-consuming and requires great expertise in the field – something that is in short supply on the labour market today.
By contrast, Microsoft Azure products can be easily integrated with other Azure products or your existing cloud infrastructure and ensure solution scalability.
IoT and Machine Learning in a Microsoft Azure cloud
Let us have a look at what Azure Cloud offers in the context of IoT and what we can potentially gain from it. First of all, there is a service known as Azure IoT Hub, which allows you to connect to and manage any IoT device, as well as ensuring easy integration.
Another service, IoT Central, allows you create usage reports, track errors and upload new app versions to your devices. This will all reduce your IoT app management and operation costs. We also have different edge computing services: Azure IoT Edge and Azure Percept, which will help you move your computing from the cloud to its edge, so that it can be performed by your IoT devices.
As for machine learning, Azure offers a range of services with wide-ranging applications. These can be divided into 3 sections: Azure Applied AI Services, Azure Cognitive Services and Azure Machine Learning.
Azure Applied AI Services
These services allow you to use AI to solve a range of simple and typical business problems, offering, e.g., chat bots, form recognition or a whole AI-based search engine. Azure Metrics Advisor will also give you recommendations about your resources or indicate potential problem areas in your infrastructure.
Azure Cognitive Services
Azure Cognitive Services offer APIs with functions such as object recognition, text translation, text-to-speech and speech-to-text conversion, Q&A, speaker intention detection and many more. To use the API, you do not need to have the complex knowledge that went into its development or even understand how it works. All you need to do is implement the integration in your app and you will be able to benefit from its analytics.
Azure Machine Learning
If none of the above services suits your needs, Azure also offers an environment in which you can develop your own model, especially in BI or Big Data: individual cases are so specialised that there can be no one-size-fit-all tool. However, Azure Machine Learning provides a simple, graphic interface that allows you to “glue together” the logic of data processing and machine learning, and easily integrate them with other Azure services, such as Storage Accounts or Azure DataFactory.
Azure IoT and Azure Cognitive Services in the transport industry: a case study
Take, for example, a transport company with a head office and a fleet of several hundred vehicles transporting their cargo in standard semitrailers, ice boxes and tanks. At the outset, all its processes were manual: orders were accepted and problems on the road were communicated via phone calls between the head office and the drivers. Transport documents were paper-based, and the drivers had to carry them at all times.
The following Azure IoT and Azure Cognitive Services were recommended:
Measure driving parameters
The fleet was fitted with a set of sensors to measure the most important driving parameters, such as position, velocity, outdoor temperature, temperature inside the ice box, pressure inside the tank, truck alerts (e.g. engine failure, low tire pressure, falling gas level).
Mobile app connection
An app was developed and installed on drivers’ phones, with a range of functions that facilitated driver communication with the head office, such as:
- reporting road accidents,
- accepting orders,
- storing transport documents,
- scanning transport documents,
- reporting arrivals/new departures.
Interconnection via Azure IoT Hub
All sensors and driver apps were interconnected via Azure IoT Hub. Together, the sensors and the phone apps that showed the location of the nearest truck created a virtual vehicle model with important data on order status.
The head office was able to easily monitor the location of every truck.
- Additional algorithms used real-time truck movement data to identify the trucks that were likely to deliver the goods behind schedule and flag them as high-risk. The operator was immediately alerted and could notify the client of a possible delay.
- Thanks to data about truck status in the event of a road accident, engine or ice box failure, or any other emergency reported by the driver, the operator was able to find the closest loading/unloading station to replace the truck with another vehicle.
This changes allowed the company to achieve the following goals
- Less time spent on communication between the head office and the drivers,
- The risk of human error is minimised throughout the process,
- The risk of delays and associated penalties is reduced,
- Sensors improve driver and cargo safety,
- All trucks can be monitored at the same time,
- The system was modernised and opened up to new investment opportunities.
Cloud computing in IoT and AI projects – conclusion
IoT and Machine Learning are fairly recent arrivals on the market. However, they are getting more and more popular and, thanks to ready-made products (available, e.g., from Microsoft Azure), you can already introduce them into your IT projects without any initial investment or expertise.
Still, system security is non-negotiable; if you fail to address it, the consequences may be catastrophic. It is always a good idea to aim for best could services delivered by specialists who will take your projects through the process and take care of the details at every stage of implementation.