Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - Teraki (Making Big Data Small for the Internet of Things)

Teaser

The data revolution is progressively arriving to the vehicles we drive in our daily routine. Our cars (even the old ones) are equipped with a large set of sensors which generate huge amounts of information regarding different parameters such as engine performance, location or...

Summary

The data revolution is progressively arriving to the vehicles we drive in our daily routine. Our cars (even the old ones) are equipped with a large set of sensors which generate huge amounts of information regarding different parameters such as engine performance, location or driver behaviour. At the same time, the Internet of Things (IoT) trend, is gaining traction across transportation industries. The result of thousands of in-vehicle sensors and IoT is the connected car or connected vehicle movement. This technology, if widely adopted, could revolutionise mobility in terms of road safety and traffic management, while having major environmental impact, representing a big step forward for society.
However, the services related to connected vehicle trends, and later on the autonomous vehicle, will generate a huge amount of information which should be send over the air to provide new services such as: predictive maintenance, real-time traffic and fleet management, usage-based insurance or even autonomous driving. The total amount of data can potentially reach 1300 Billion GB per year, which will largely overcome current cellular network capacity, making this services non-feasible at large scale. Moreover, even if some of this services are available through new communication standards such as 4G-LTE or 5G, sending such amount of information will be prohibitively expensive. At TERAKI, we have developed a breakthrough compression algorithm that will be the key enabling technology for new business models on the connected vehicle industry. We achieve and average data reduction rate of 90% on a wide range of in-vehicle sensors (thus, sending 10 times less information), while reaching a reconstruction accuracy of 99%. This will allow us to unlock the provision of new data-intensive connected services while extremely reducing the cost of existing ones.
With the present project, we aim at introducing improvements on our current solution that will allow us to increase even more the reduction rate and the accuracy, while building a scalable architecture ready for a mass-market adoption. After analysing both the market opportunities and the technical feasibility we decided to apply for the phase 2.

Work performed

During the phase 1, we have carried out mainly the following tasks: 1) Market and business plan assessment, 2) Validate assumptions with our customers and analyse risks, 3) Technical and logistic feasibility, 4) Certification and regulatory analysis, 5) Intellectual property study and 6) we build a Detailed work plan for Phase 2 application.
We have defined the main objectives to be fulfilled to convert our prototype into a product and also defined a detailed technical roadmap to ensure a proper execution of all the tasks. We confirmed that we have the internal capabilities to successfully implement all the envisaged developments except: High performance computing and code parallelization, high scalability of backend systems and development of SDKs microcontrollers. To cover this needs we looked for the best subcontractor in this regard.
From the other side, we corroborated our freedom to operate and planned our future protection strategy as well as assessed the main certifications/standards we need to fulfil to sell our product. Finally, after analysing the market conditions of several IoT verticals, we concluded that the transportation segment is the most adequate to fit our solution, in particular the “connected vehicle” which fits under the road transportation sub-segment. We confirmed the hypothesis by having direct interviews with 2 of the largest automotive manufacturers in Germany.
Based on our potential customer’s pains and requirements, we also defined the business model and pricing strategy that better suits them and, at the same time, maximizes our margin and revenues.
After corroborating that our expansion and implementation plans are feasible, we decided to apply for a phase 2 under the SMEI.

Final results

IoT services for road transport vehicles including aftermarket telematics units, connected cars and autonomous driving will be a 60 Billion industry by 2020, bringing to the roads approximately 70 Millions of connected vehicles. Moreover, the amount of devices will be exponentially growing until 2040. As mentioned earlier, data-driven services related to these segments are prone to potentially send more than 3600 Million GB per day, or what is the same, around 1300 Billion GB per year.

This context poses the right premises to sell compression technologies that help dealing with this vast amount of data. Several vertical solutions have appeared over the years to solve particular problems such as codecs and intelligent algorithms for compressing text (GZIP), images (JPEG) videos (QuickTime, Alpary) or music (MP3). However, in the connected vehicle domain, the information is extremely heterogeneous since is generated by a wide range of sensors capturing completely different physical magnitudes: acceleration, angular rotation, temperature, velocity, time, GPS coordinates, object distances, etc… Therefore, the sector is requesting a transversal and generic solution able to automatically adapt to each type of signal constraints in order to unlock data-driven services. However current solutions that aim at solving this problem have several drawbacks:
- Filtering techniques require high CPU usage, which makes them not suitable for embedded systems mounted within a car. Moreover, unforeseen events, represented by frequencies that are cut off by the filter, cannot be reproduced, so that relevant information might be lost and thus, the accuracy highly decrease
- Subsampling techniques are not versatile enough since they offer only 2 options: 1) Have a low variation on signal reconstruction but at low compression rates or 2) Have high error on the reconstruction but achieving high compression rates.
- Generic compression techniques (e.g. zlib, glib) are intended to solve only specific data-problems (text, image, etc…), requires high CPU usage and obtains limited compression rates and accuracy when applied to “generic” signals.
However, TERAKI is the answer to all this limitations, since it is a versatile solution capable of compressing any kind of vehicle sensor data with high performance.
Only considering the amount of data to be sent by connected cars in 2020 on our targeted segments, we are targeting a potential business opportunity of more than € 19 Billion in 2020. During the phase 1, we have defined a hybrid business model that will allow us to capture a significant share of this huge market, in short:
1. Virtual telecommunication operator (MVNE) In this scenario, we will act as the operator responsible of the whole process of data gathering, storage and reconstruction in our external servers (including the SIM card provision). We forecast that we will offer a total pricing 30% lower than current offerings.
2. Licensing, we will charge monthly fees to licence our technology to industry customers.

Through these hybrid models, we will sell our services to the most relevant players on the sector. We expect that this will allow us to generate a yearly revenue of €32M already in 2022, generating 60 new jobs in Europe, becoming the intermediate telecom operator for the connected vehicle industry.
TERAKI will unlock new data-driven services associated to the connected car. Leaving apart, the economical outcomes, TERAKI will have high impact on EU society. Just to provide some examples: collision avoidance systems could reduce fatalities by 30% in Europe, with a further 30% reduction in serious injuries. It is clear that if 2 connected cars were communicating to each other in real-time, they could avoid a frontal collision much in advance we can even notice it. From the other side, real-time vehicle location data combined with simulation models to build predictive traffic information has the potential to reduce congestion by 15%, which is extremely relevant considering that it costs EU nearly €100 billion annually.