IoT News, Publications, Magazine & Newsletter | IoT Now

As the automotive industry gears up for self-driving and connected cars technology transfer into the market from other areas is accelerating innovation. Neural networks, sensor fusion, robust networking options and heightened security will enable connected cars to offer more automation, Lars Reger and Juergen Weyer tell IoT Now.

IoT Now: To what extent do you see technology developed in the automotive sector being transferred to smart grid deployments? There’s clear synergy between vehicular activity and the smart cities concept but do you feel technologies can be complementary?

Lars Reger: If you look at the high level concept of a self driving robot, which is where the automotive sector is heading, and a device that senses its environment and sends a command to the actuators of a home environment they are, from an architectural point of view, super-similar. The process of sensing, thinking, acting is common whether in a smart meter or a car. Each involves calculations being performed and a simple set of instructions being sent. In cars, those are accelerate, decelerate or change direction; while in a smart meter it’s switching a device so they’re pretty similar.

In both devices there is the paranoia about errors being made so you have to ensure they are failsafe, but also secure so they can’t be hacked. Therefore you need data security but a lot of it is simple from a chip point of view.

Juergen Weyer: It has to be real-time, safe and secure, protected and always available. If it’s not, the consequences will be that the consumer has no energy arriving in the smart home. In the automotive environment, users won’t be able to travel.

IoT Now: Well publicised hacks of in-vehicle systems, such as the Wired journalist’s Jeep hack, and the generally high consequences of security breaches for connected cars have made vehicle security a high priority. Do you think automotive security and safety technology now leads other sectors and what lessons can be learnt from the automotive market?

LR: You could create extreme dark use cases on the road but also on the grid. Both, therefore, require zero tolerance for failure. Automotive is in the middle of the security discussion but the car industry is not leading the initiatives. NXP is the second largest supplier in the automotive sector, but we are also the biggest crypto provider worldwide. We have taken use cases and intellectual property from our sister division in NXP and that has enabled us to jump ahead. Without that, we’d have lost ten years in development and standardisation.

JW: The issue of security in cars has not just happened with the recent hacking reports. We’ve been looking to make cars secure with immobilisation and to prevent mileage frauds for many years. With our heritage from Freescale we have a lot of experience of security across secure networks, having provided high-end multicore systems, and there are lots of security mechanisms in network architectures that have been included in network processors.

It’s interesting that some technology comes from automotive but is enriched by our crypto and network capabilities.

IoT Now: The stakes are equally high when it comes to unmanned aerial vehicles (UAVs). What can other sectors learn here and what do you see as the main attributes being developed in the aerospace sector?

LR: We’ve been asked to enable UAV-to-UAV communications because under normal aviation law UAVs are mandated to have anticollision transponders. Those developed by Boeing and Airbus for use in jets typically weigh about 10kg, so there’s an obvious issue if you want to use a UAV to deliver a 3kg parcel because you’ll have to lift and fly about 14kg to do that.

It’s also important that a UAV is able to identify itself. NXP has utilised technology it originally developed to enable ambulances to switch traffic lights to green as they approach lights. Using the same electronics, a UAV filming a large event can tell anyone that it is a CNN TV UAV allowed to be present. Identification is important because everyone can see the value of knowing a UAV is an Amazon UAV carrying a 3kg delivery rather than a UAV carrying a 3kg bomb.

We have very tiny radar modules which are also being used for anti-collision systems and enabling UAVs to land between buildings and homes. We also have motion detectors which can protect children and pets from the rotor blades of UAVs which are 40-80 cm long and razor sharp. No one wants those near their child or dog.

We are therefore lifting innovation from other areas into UAVs. For example, we’re working on UAV licensing. Today, if I crashed my UAV I could walk away and no one would know I was the pilot. However, authorities are starting to bring in licensing and we have an NFC-based solution that involves a certificate in a mobile phone which the UAV must read before taking off. This means different pilots can fly UAVs and the user is always identified. We think this sort of regulation will be more in demand as UAV usage proliferates.

IoT Now: Much innovation is being seen in the area of sensor fusion. What developments are you seeing and how might these find applications in the automotive sector?

JW: When we look at the evolution of cars and the innovation from electronics we look at what we call driver systems such as adaptive cruise control, lane departure assist or brake assist. All of these systems help the driver in situations where they are distracted such as when taking a phone call. However, when we get to the next level of automation – which will be highly automated, not autonomous – something has to go into the car that replaces our human senses of seeing and hearing. We, as humans, have natural sensor fusion but cars will need to combine radar with vision systems such as cameras and lasers. The interesting part comes in bringing all of that data together to a microcontroller with the last part being to make that more intelligent.

The real challenge will be what systems learn to trust and how to fuse all that information to make it descriptive and to give inputs to a real driving situation. We have to bring small technologies from many different areas together to achieve this.

LR: The goal is to build a machine that is always superior to the human capability. Radar, for example, is a system that never gets blinded by sunlight and can look through fog. The only requirement for developers is not to artificially narrow the view or blind the system, allowing it to misinterpret the situation. The value of sensor fusion comes through taking inputs from a variety of sensors and fusing that before analysing it to ensure a consistent picture is being provided. We’re providing the platform for all of that.

IoT Now: As 5G connectivity continues development how do you see it co-existing with 802.11p? Do you expect services to rely on access to both technologies?

LR: Today you have cellular and 802.11p, which is a peer-to-peer straight communication. There’s a good need for both. For example, if you’re driving from Hamburg to Munich the road goes through several tunnels and areas of no coverage, but for traffic system updates, cellular is fine. It’s not time-critical and connects to the web to gather the data.

However, for applications such as a project in the Netherlands that involves two trucks following each other closely on a virtual towbar at 80 km/h at a distance of 10 metres, the brakes of the following truck are required to be activated within 4ms of the truck in front braking.

This is information that, of course, cannot be delayed or interrupted and the fastest connection is direct from peer-to-peer. Peerto-peer is the only way this application can work because a truck-to base station-to truck communication involves too much delay because of the latency of the mobile network.

A peer-to-peer LTE standard exists but today it involves asking if another truck is permitted to communicate so it wouldn’t work for that close-running braking application. 802.11p is the only technology robust enough.

JW: We’re talking here about safety critical, life-ensuring demands. The latency of information in a cellular network is one aspect and the availability is another. I don’t think we can have safety impacting information delivered using cellular networks. We still get dropped calls and what is required for these applications is 100% availability right now, to the second. Cellular communications can’t do that so we will have co-existing technologies.

IoT Now: What role do you see artificial intelligence playing in automotive applications?

LR: There is huge interest and it is centred around neural networks and self-learning for sensor fusion boxes. Neural networks can do recognition jobs quite well because the sensory information enables them to locate objects. Cars, while driving, are getting data using fixed recognition algorithms and makers such as BMW or GM or Toyota are using some form of AI to report back on that data. The industry in general is learning how neural networks should be programmed.

JW: We still have to recognise we’re in an automotive environment that has very stringent regulatory and performance requirements. We’re providing the industry with robust technologies for the 15 year relationship of a car and its driver that can work at extreme temperatures. If you don’t address these requirements at the design stage you can’t get to the level required.