NEXYAD Automotive & Transportation Newsletter #29, March 24th, 2020

Headlines :

– Nexyad CEO Editorial

– CTS 2020 (Consumer Telematics Show)

– CES 2020 (Consumer Electronics Show)

– Nexyad Quoted in a CNN Article

– Global Road Safety Platform by Nexyad

– Deep Learning Benchmark : Tensor Flow Vs Nexyad Internal Deep Learning Solution

NEXYAD newsletter: our Global Road Safety Platform                                               

Hello,

In these difficult times of COVID19 epidemic, I would first like to wish everyone good health, for themselves and their loved ones. Our company NEXYAD continues its activity, with difficulty because organization of telework in a software development company is a complex security problem, but we achieve the planned deliveries. All our customers still can count on us.

Our company is about to open a new page of its development, with the achievement of our Glogal Road Safety Platform, presented in this newsletter, which allows our customers to develop their own mobility solutions.

Applications of this NEXYAD Global Road Safety Platform are mostly:

1. Telematics: alerting driver when driving risk is too high avoids 20% of road accidents, which represents a very significant Return On Investment (ROI): 200€ per year per vehicle.
   . for fleets: Risk Management
   . for insurers: Usage Based Inssurance (UBI)
   
   2. Automotive:
   . we know how to calculate the speed of the vehicle and its acceleration, at all times, which guarantee to stay below an acceptable maximum level of risk. Those car speed and acceleration are the typical inputs of an Automatic Cruise Control (ACC) system;
   . we know how to inform the autonomous vehicle of the risk that it is taking, in real time, Thus allowing it, on the one hand can be used to modulate its behaviour by adapting to new and unknown situations, and on the other hand, to record situations where the risk has risen too high in order to add them to the deep learning databases. NEXYAD technology is a good lead to go from AD systems level 3 to level 5 without increasing the number of sensors and the computing power: human driver doesn’t have all those sensors and can drive safely in complex situations. We propose the same scheme based on anticipation.

Our technology is available for cars, trucks, and motorcycles. This technology is now starting its massive deployment in Europe, USA, Asia and India. We are proud of our customers’ products and solutions that integrate, among other things, our technology. 

We Save Lives

This year, for the first time, Nexyad has attented the Consumer Telematics Show January 6th in Planet Hollywood Resort & Casino, Las Vegas. 

We followed several interesting speeches and round tables during a full day conference with speakers from Roborace, ZF Group, Nissan North America, Veoneer, Ford Motor Company, Karma Automotive, Volvo Car Group, Avis Budget Group, Toyota North America, LexisNexis Risk Solutions, Audi, Amazon Web services, and many others.

Finnish companies were well represented on the booths corner :

Vaisala enhances safety, efficiency and decision making through environmental measurement and related services.

Tuxera creates quality-assured software to help to store and do more with data.

Forciot develops advanced IoT sensor solutions for  automotive, logistics and wearable technology sectors.

Flexound Augmented Audio™ adds the sensation of touch to audio-visual listening experience.

Unikie founded in 2015 is a company in software industry that provides  unmatched service and competence.

Vincit designs and engineer software, provides digital services, and products.

Note: the presence of car maker KARMA with a beautiful e-sport sedan (see below).

* * * * *

View of Central Plaza at LVCC

For the third consecutive time, Nexyad took his quarter at Las Vegas Convention Center, Central Plaza into French Fab Pavillon just beside Faurecia and not far from Here Technologies and Google.

Inside this French Fab Pavillon, some SMEs of high technology received lots of visitors, prospects and customers. 

Actronika develop human-machine interface solutions using patented haptic technology.
Humans are born to touch in order to form relationships, to learn, and to be engaged with environment. Faced with an ever-digitizing world, this company made it as mission to re-engage people with tangible, material experiences.
Their hardware and software make that possible.

Since 2002, Benomad have been designing, developing and marketing mapping and navigation software tailored to helping companies manage their mobile resources as effectively as possible: Business and consumer navigation application; Mapping services with simple and low cost fleet management; Integration tools like mapping SDK and Navigation API, etc.

Eldim was born in 1992, its engineering teams has acquired a high knowledge in the optical analysis of angular properties of the light in transmissive and reflective mode. They develop and manufacture: Fourier optics viewing; angle systems; video photometers, colorimeters and temporal analysis systems; viewing NIR characterization sources and turn-key inspection systems.

Since 2012, Geoflex provides accuracy, continuity and integrity on radionavigation satellites measurements. Its partnership with CNES (French national center for space studies) allowed to industrialize and commercialize the PPP-CNES technology consisting to model and estimate in real time all the different errors affecting GNSS measurements, in order to add value in the fields of positioning and navigation, precise timing and meteorological forecast model.

Prove & Run was created with the idea that large-scale deployments of connected objects is an attractive target for remote cyberattacks and that the solutions used in the mobile industry are not sufficient to answer to the challenge. They provide cost-effective, highly secure, off-the-shelf TEEs and hypervisors that dramatically improve the level of security of connected systems.

SBG Systems designs, manufactures and market a complete line of inertial sensors based on the state-of-the-art MEMS technology such as Attitude and Heading Reference System (AHRS), Inertial Measurement Unit (IMU), Inertial Navigation Systems with embedded GPS (INS/GPS). They address, for more than 10 years, the most exciting markets.

Stemming from 15 years of know-how in connected, cooperative & autonomous mobility YoGoKo is an industry-leading communication solutions provider for the connected, cooperative & autonomous vehicles (CCAV) evolving in intelligent environments. For transport & mobility market players: autonomous vehicle manufacturers, legacy automotive OEM and Tier1, for intelligent mobility solutions developer and integrators.

Inside the big Automotive cluster MOV’EO, there is another cluster called Groupement ADAS obviously specialized in the field of Advanced Driver Assistance Systems and Autonomous Vehicles. 13 companies currently compose it, and this year Intempora and Nexyad were animators to explain synergie, skills, services and products providing by more than 200 engineers, doctors, professors and experts.

When a French SME wants to have a talk with French Secretary of State to the Minister of Economy and Finance, Ms. Agnès Pannier-Runacher, no better idea than travels 5400 miles going at CES in Nevada USA.

Agnès PANNIER-RUNACHER talked with NEXYAD CEO Gérard YAHIAOUI

Nexyad has met American company VSI Labs with the goal to work together soonly. They are specialized in vehicle demo on open road or test road for ADAS and Autonomous Vehicle technologies on the sidelines of congresses and shows. They mainly work in USA at the moment and they plan to organize sessions in Europe and Asia in near future. We plan to build a partnership with VSI Labs to show our global road safety platform to American market.

Next to the French Fab Pavillon, three big suppliers in Automotive have put their booth, Valeo, Here Technologies and Faurecia. These giants come to meet their customers and providers like Nexyad.

In North Hall, there are almost all the big car makers, the engineering companies, and Tear ones in luxurious spaces.

Some most impressive machines for next generation mobility:

Nexyad talked a lot with attendees, organisations and partners and many thought this year shows a little decline of shuttles and autonomous mobility exhibitors. For us, it could be the proof that market is becoming mature. At the same time, full autonomy technologies for road vehicles is postpone year after year. The safety asked by authorities and public fails yet to be demonstrate. We think this issue can be overcome if engineers worldwide considered driving risk as an explicit variable instead of an implicit one.

* * * * *

Nexyad Quoted in a CNN Article

“Derq is not the only company working in this space. Humanising Autonomy has developed a “pedestrian intent prediction platform” intended to make autonomous vehicles safer;” Driveri is a platform that analyzes commercial vehicle drivers’ to improve their performance, and Nexyad creates software that can be built into cars to alert drivers of potential accidents.”

Click to read entire article

* * * * *

Here is the presentation of our global road safety platform: computing Driving risk taken by vehicle in its Driving context, 20 times per second allows to:

. act on human driver through risk alerts (accident rate reduction by 20% at least, validated by road safety experts and insurers).

. act on automated vehicle: Automatic Cruise Control desired speed and accels / Autonomous Driving system that is aware of the Driving risk it is currently taking: being servo-control to keep risk under a max accepted value, and triggering use case recording for next deep Learning versions.

* * * * *

Several times per year, we benchmark our internal Deep Learning solution (IP NEXYAD) (NxDev) vs Tensor Flow (Google) that is a very efficient Learning system. Results again this time (Feb2020) is that NxDev still beats Tensor Flow. Of course we are happy about it, and we still work (since 1995) on Learning efficiency improvement, But comparing 2 deep Learning systems is not that easy.

First : we select a database (the same for both) and a function. This time : classification.

Second : we choose a performance measurement. Many people use % of good classification, but we chose Khi-2 value.

Third : train with exactly the same number of neurons, layers … and on the same database.

Fourth : compare Khi-2 values.

We made iterations with a growing number of neurons to draw the Vapnik curves – check if a better performance couldn’t happen with Tensor Flow for a different architecture.

We also Added perturbations with mislabeled added Learning examples, and restarted the whole process.

If other teams already worked on Learning systems benchmarks (including Google team), we would be happy to share skills and improve comparison method.