Top Seven Use Cases for Streaming IoT Applications in the Automotive Industry

IoT networks in conjunction with machine learning are reshaping the automotive industry for manufacturers, drivers, and maintenance technicians alike. More than 250 million vehicles are already equipped with IoT, now sometimes called IoV – Internet of Vehicles. IoT sensors in factories detect parts inventory levels and reorder to keep production flowing smoothly. They use computer vision to predict and detect defects in production. Sensors detect driver vital signs and make life-saving recommendations including rest stops. They also warn of impending maintenance issues by sensing heat, vibration, and other wear-and-tear, ultimately preventing dangerous breakdowns on the road. 

The benefits of IoT in the automotive sector are driving extreme interest and growth. Data throughput from sensors and actuators will require stream processing on an exabyte scale, so much so that the now optional Edge Compute paradigm will soon become imperative. This is true for two important reasons. 1) IoT networks will produce so much data that it’s not effective to send it all to Cloud processing. 2) Edge ML app will extract intelligence near the data source, so that only analytics and BI outcomes are sent efficiently to the Cloud. Even with Edge optimization, performance will depend on the fastest, most accurate stream processing, especially in Mobile Edge Compute (MEC) scenarios via the best stream processing platform, which in this broad use case is Apache Pulsar.  

Our purpose here is to define the emerging synthesis of IoT automotive components across manufacturing, driving and maintenance:

• IoT – IoV sensor data
• Live event streaming
• Machine learning applications 
• Mobile edge computing

These new IoT equipped vehicles enable us to realize a connectivity and interaction between physical environment and people which augments awareness of danger and improves driver experience. Much of the same intelligence which autonomous vehicles derive from the environment will now be available to human drivers to enhance efficiency, safety and comfort. 

In automotive use cases, IoT devices will improve transportation efficiency  by integrating with existing Intelligent Transportation Systems (ITS) via infrastructure, including 4G and 5G, and to be always online and always streaming updates to deep neural networks. Intelligence from ITS will reduce travel delays and disruptions due to accidents, construction, damaged roads, weather hazards and much more. Driver and traffic social networking apps, in which drivers can chat with each other to share information about jams and delays may also be mined for data in parallel with IoT to improve awareness of environmental conditions. 

IoT-equipped vehicles can participate in real-time vehicular performance ratings, monitoring of emissions, early reporting in diagnosis of mechanical failures, highway traffic analysis, and contribute to smart automotive manufacturing. The automotive IoT ecosystem is highly inclusive and encourages data and intelligence sharing across all stakeholders.

Automotive IoT ecosystem components:

Smart Manufacturing

• Supply chain  
• Materials and parts
• Production optimization

Driver Experience

• Real-time navigation
• Traffic management 
• IoT enhanced black box 
• Emergency vehicle dispatch
• Driver vital signs monitoring
• Safety issue forecasting
• Infotainment systems or driver dashboard smart panels and displays

Maintenance and Service

• Unscheduled maintenance
• Core parts supply/demand
• Testing and diagnostics
• Quality assurance
• Service personnel vital signs monitoring

Sensing devices such as gas and oil level, speed and odometer, engine and radiator temperature, even tire pressure were always widely used in vehicles; however, when those same instruments are connected to LAN and internet enabled as IoT, they now feed holistic machine learning algorithms which vastly enhance awareness and forecasting of future issues. 

Moving Sources and Sinks

An important technical challenge in the growing automotive IoT field is scalability, because of the enormous amount of data potentially generated by sensors. Mobile Edge Compute (MEC) must evolve quickly to handle much of the insight extraction locally, perhaps with onboard computer systems with capabilities similar to those of autonomous cars. ML algorithms must run locally, in onboard computers, to provide instant results from ML about obstacle and hazard recognition under immediate time pressure. Onboard nodes must then compare outcomes with longer term analytics and decision support when time constraints permit (when the vehicle is not in use). 

Ideally intelligence will comprise the bulk of streaming data. Yet, this will still be a massive amount of data which must be managed by a scalable and reliable pub/sub live streaming platform. The best in class for this type of use case is Apache Pulsar. Given the nature of the car as a mobile data center and as a set of  sources and sinks which are always in transit, Pulsar’s Cloud native design features are ideally suited with:

• Separation of store and compute
• Geo-replication
• Multi-tenancy
• Zero data loss guarantee
• Lowest latency in industry

Pulsar works perfectly with the emerging IoV networks of interconnected vehicles to feed MEC networks and optimize data burden to Cloud processing with the three types of Edge Compute for automotive use cases:

• Fog computing 
• Mobile Edge Computing (MEC) 
• Cloudlets 

Fog Computing connects devices and Cloud to guarantee security and interoperability with low latency targets. MEC resolves data volume and latency problems by connecting vehicles as networks. Cloudlets then have the purpose of providing powerful computational resources to reduce data transmission volumes. Combining the three methodologies will be necessary as data to ML scales upward in new orders of magnitude. In this context, many enterprises will find benefit in partnering with a managed service to support such an infrastructure – like Pandio. Pandio’s existing knowledge base in the Automotive IoT to ML stream processing arena means that they’ve already solved problems you will encounter eventually.

With that context, let’s explore some of the most compelling use cases for streaming IoT in the automotive industry.

1. Real-time Navigation

While navigation is already optimized in the Google Cloud, wherein Google logs speed data from signed-in Android devices to visualize in online traffic congestion maps, there will be clever new purposes for this and more data. For example, imagine combining the user comments about causes of traffic jams so that a driver can say to her infotainment system, “Hey! what’s causing that jam at I-85 and Wendover Avenue?” The smart transportation network, listening via IoT to conversations in cars near the site can derive keywords such as “crash” and “police checkpoint,” or “flood.” The myriad of possibilities quickly point to a common looming challenge: scaling up to exabytes of data for ML apps. 

Until the emergence of IoV networks, drivers used navigation maps as a point of reference. But now imagine all those maps talking to each other! Imagine an IoT reporting an emergency in which all victims are unconscious and unable to call for help. The ITS receiving such a beacon will then automatically notify nearby rescue teams and even dispatch one to the accident. Additionally, ML algorithms will gradually improve their accuracy with emerging outcome data from previous prediction success and failure.

2. Smart Traffic Management 

As a promethean goal, a widely adopted cooperative IoT traffic management system can actually improve a nation’s economy by resolving problems in operations research, the logistics of transportation flow. An ideal task for ML deep neural networks, the intelligent syncing of traffic signals as learned by training data on weather conditions, vehicle density, and other factors can accelerate the whole network substantially. ITS can then alert drivers about upcoming conditions, suggest parking intelligently, warn of upcoming toll bridges, floods, and much more. An ITS extracting intel from MEC networks may deduce vehicle physical characteristics and update suggested routing logistics algorithms; the system may learn entirely new patterns in operations research to suggest novel improvements. 

3. Enhanced Black Box 

The event data recorder known as the Black Box commonly used in airplanes will be transformative as a natural IoT device, and will contribute data to investigative agencies. The problem of a traditional recorder being destroyed or damaged in an accident is solved by an IoT Black Box transmitting critical data as a real-time stream of events. An IoT-enabled black box will likely cause the re-invention of monitoring and surveillance to essentially eliminate doubt with regard to responsibility or liability in traffic accidents. There will be no need to find the recorder after an incident, nor any concern that it was destroyed in a crash, because the data was streamed up to and possibly beyond the point of disaster. With Apache Pulsar’s zero data loss streaming messaging platform, the critical data and decision support will survive the incident.

4. Emergency Vehicle Dispatch 

We briefly covered emergency vehicle (EV) dispatch optimization above in the section on IoT enhanced navigation. The optimal routing of EVs is obvious, but there is a particular focus of interest to add. We have also touched on the concept of IoT monitoring vital signs of drivers, and this is especially important in the diagnostic assistance potential of IoT streaming to ML medical instruments and https://www.mcmedicalnj.com apps. In particular, when onboard IoT devices may transmit intel about a potential cardiac arrest victim, the fact may be used to stimulate a faster or higher priority EV dispatch.

The reasoning is based on research which suggests that even a difference of 60 seconds response time can determine the survival of a heart attack victim. Here is a medical emergency scope for IoT and ML apps in automotive use cases which justifies considerable exploration and development. Here again, domain knowledge expert partners such as Pandio who have already built open source libraries to drive ML training and deployment, represents a positive step in the direction of success in such projects.

5. Pollution monitoring

According to WHO, there are about  2.4 million casualties annually as a direct result of air pollution. Air pollution is a critically hazardous condition which is a prime target for monitoring and alert by IoT sensors coupled with ML and AI. Volatile gasses, greenhouse gasses, and other toxic chemical compounds are safety concerns which will be managed more effectively when monitored by automated networks of IoT devices.

6. Driver well-being check

IoT sensors will likely supply a bevvy of data about a driver’s physical and even psychological conditions. Sensor data can provide, for example: 

• Electroencephalogram (EEG)
• Electromyogram (EMG)
• Electrocardiogram (ECG)
• Heart rate 

The most obvious targets are alcohol, DUI and immediate risks which would endanger driver and passengers. As we traverse these individual use cases, a compelling narrative evolves: the collection of sensors provide data which together trains ML models to integrate toward increasingly holistic views. Shouting, increased speed, elevated heart rate, all suggesting intervention.

On a more typical analytical note, standard measurements of rising engine temperature, rapid fuel consumption, commensurate exhaust emission, lead to a warning about the turbocharger compressor and a seized engine.

7. Smart Supply Chain Management 

IoT streaming ML apps in the automotive industry have the potential to improve information sharing among manufacturers and their suppliers. Cooperation at this level will then benefit consumers. For example, when a manufacturer shares ML forecasted vehicle demand projections with inventory  suppliers, now the suppliers can match orders exactly and avoid surplus and shortage conditions. Uninterrupted production likewise benefits seller and consumer plans.

IoT ML Streaming Experts

Architectures for streaming real-time event data from IoT to AI and machine learning apps can certainly appear daunting in complexity. At the highest design level, for example, which apps in the following component categories integrate seamlessly from one to the next to reach the goal of extracting intelligence? 

• IoT devices and hub

• IoT open source platform
• Streaming messaging
• Edge compute framework
• Database and query
• AI – ML algorithms
• BI – analytics outcomes

A proven best-in-class combination of the above includes the ideal Cloud native streaming messaging technology Apache Pulsar. The fastest, most reliable, secure, lightweight framework for Edge computing, Pulsar repeatedly beats Kafka in performance benchmarks. Plugging IoT into the Edge compute paradigm, Pandio excels with expertise in PrestoDB to source training models within its own PandioML deep learning library.

A managed service like Pandio, which can seamlessly integrate all these components will inevitably save expense over staffing in-house development teams. In other words, Pandio already has an integrated open source solution with optimized Pulsar and Presto configurations for streaming IoT to ML in automotive apps to support decision making. Hooking your concept into Pandio’s managed service speeds time to market while minimizing in-house technical complexity.