TelcoFuturism - Part 4
Connected & Autonomous Vehicles
This is the fourth blog post in a series on “TelcoFuturism” – looking at the points of intersection between the telecom industry, areas of advanced technology, such as drones, AI, AR/VR and robotics, plus societal changes such as climate change & shifting demographics. Disruptive Analysis looks at these adjacent technologies through the lens of what’s really happening in networking, opportunities and challenges that could emerge and the tough practicalities and complexities telcos face, rather than accepting the general hype and rose-tinted views common among some forecasters.
Background: Don’t focus on the ultimate endpoint
The last few years have seen a massive growth of interest in autonomous vehicles, with near-daily visions of fully driverless vehicles. We have seen YouTubers filming themselves in the back seats of their own cars, with nobody at the controls. We have witnessed numerous camera- and sensor-laden test vehicles driving around cities and along highways
And we have, unfortunately, seen the results of unlikely (but not impossible) “corner cases” where pedestrian or passenger casualties have resulted from imperfect and immature technology.
There have also been numerous companies and policymakers across the telecoms industry talk about self-driving vehicles as a key use-case for 5G networks’ ultra-low latency abilities, network slicing or perhaps the need for ubiquitous edge-computing near to roadsides.
To Disruptive Analysis, much of this sounds like hype. 5G doesn’t need AVs, and AVs don’t need 5G. It will help, but it won’t be a pre-requisite … because we will never have ubiquitous 5G coverage and unlimited capacity, so vehicles will also need to work safely when offline, or in partial or degraded coverage.
This blog post isn’t about a full analysis of the state of the art of full autonomous vehicle technology, or indeed 5G radio limitations. Instead, it focuses on what is realistic over the next 5 years and a few years beyond that. And for that, the key term is not AV but CAV – Connected & Autonomous Vehicles.
What will happen – indeed, is already happening – is that CAVs are being connected to both public mobile networks, and to localised networks run by government authorities and highway agencies.
There are numerous use-cases and architectures emerging (and we covered connected highways in a previous post), but this article focuses on the telecom/network implications of:
- The combination of electric vehicles (EVs) and CAV technology
- “Assisted driving” rather than “Autonomous driving”
- Local breakout and edge-interconnect
- Non-car use cases of autonomous driving
- The changing landscape of cities in the AV era
EVs + connectivity
The transition of vehicles from internal-combustion engines to electric propulsion brings a variety of new use-cases for connectivity, both in terms of real-time control and offline/delayed data upload and analysis.
Some of the most interesting intersection points of EVs, CAVs and telecoms include:
- Connected charging stations, for multiple purposes such as identity and payment, bulk data upload / software download via wired connection, or even advertising and entertainment screens, while the driver is waiting for a fast-charge to occur. Providing services for EV charging networks is a significant opportunity for telcos over the next few years.
- Battery management has various angles, including the ability for a cloud-based service to optimise energy consumption, protect battery health and optimise range/demand for charging stations. While most of this doesn’t require fast or low-latency connections, there are some use-cases that allow fine-grained adjustment to power/throttle levels based on surrounding traffic and other factors. (For instance – a car can be allowed to coast over the brow of a hill, if it needs to brake for traffic lights shortly after anyway)
- City or nationwide power management could allow broad fleets of electric vehicles’ charging or usage to fit better with national targets for CO2 emissions. That could be for municipal public transport, or perhaps even for adjusting consumers’ vehicles or charging differentially for their use. Is “Ludicrous Mode” on a Tesla really justifiable with an imminent net-zero carbon target?
Long before we get to fully autonomy, we are already seeing vehicles with “assisted” driving – cars using sensors & onboard processing, to keep in their lane on a highway, automate parallel parking, and brake more quickly in an emergency. Numerous other innovations are occurring in this field – and the sensor data is also used to improve the algorithms for future autonomy.
But these are almost entirely performed within the vehicle. The sensors, compute platform and actuators don’t rely on external sources of data or decision-making. A separate set of assistance applications involves external connectivity to the vehicle – either from other vehicles (V2V) or from roadside infrastructure (V2I).
These technologies (collectively known as V2X) are also maturing rapidly, although given the safety implications there remains a lot of conservatism among both authorities and auto-makers. The main focus has been on specific, clear use-cases that indisputably improve outcomes, without risks of negative unintended consequences.
A good example of this is the inclusion of eCall systems in new European vehicles, which automatically initiate a call to emergency services, in the event of detecting a serious crash. They can communicate location, direction of travel and other data, even if a driver or passenger is injured or incapacitated, and unable to call on their own phones.
The next steps in this world are under investigation and early deployment, using both cellular technologies and others such as DSRC (dedicated short-range communications). Various models are possible, either with the RSUs (roadside units) owned and operated by government agencies, or in combination with traditional mobile operators
This is where low-latency starts to become particularly important – for example, if sensors at a junction detect someone jumping red lights, and broadcast a signal to all other vehicles near the intersection to be aware, or brake. Initially, such deployments are likely to be localized to specific danger spots, or perhaps across central business districts of advanced cities.
More sophisticated models could require much broader network coverage – for example black-ice detection along a full road network, fed by traction-loss data from some vehicles, and transmitted to others approaching that location.
Local breakout & edge-interconnect
Many in the telecoms industry see the CAV market as ripe for the adoption of edge-computing services, perhaps linked to ultra-low latency (URLLC) 5G. The idea is that vehicles’ sensor data can be transmitted, processed and acted upon within milliseconds. Multiple vehicles could be coordinated into “platoons”, and road conditions could instantly be reported and used to change driving patterns. Video feeds could be correlated, filtered or combined. Traffic-safety information could be stored and retrieved locally, rather than being transported to centralized data centres.
The idea is that the compute nodes could be closely integrated into the cellular infrastructure – perhaps at aggregation point, or even co-located at the individual cell-towers and roadside units.
Yet while this is an appealing vision, it also poses some challenges. Vehicles and roadside infrastructure is likely to rely on a blend of public mobile operators, government agencies, and perhaps third-party infrastructure from neutral-host companies or private landlords. Each may have their own edge-compute and transport network resources.
While each may individually provide useful functionality, many of the use-cases will need them to be combined. If Vehicle A (connected with 5G Operator 1) and Vehicle B (connected with 5G Operator 2) are about to collide at a blind bend, how does this data get collected and acted upon, perhaps where the roadside units are run by Operator 3?
Where is the “edge interconnect”? How do the different network operators’ edge capabilities share data, and where? It obviously makes no sense to go all the way back to the core network – but there is little scope for “local peering” in today’s architecture. Indeed, very few networks have “local breakout” of data from the telco domain to the nearest Internet node or general-purpose data centre.
For many of the suggested CAV use-cases, this will need to change, especially for low-latency safety issues. There will need to be more interconnect points, more offload/breakout, more local connective fibre and more shared “neutral edge” facilities. Some of this can be facilitated by CUPS (control and user-plane separation) in the 5G core system, but it doesn’t solve the overall issue.
This is a big potential opportunity for innovative infrastructure providers.
Non-car uses of autonomous driving
When many in the telecoms industry think about CAVs, they (understandably) focus on passenger cars, whether those are owned by individuals, or by ride-sharing and rental companies. There is an implicit belief that these will help drive the need for 5G, URLLC connections and perhaps network-slicing over national networks as a whole.
Disruptive Analysis thinks that may be true in the very long term (7+ years), but for now, there is a much greater chance that full autonomy will appear first in other contexts. Initially, the action will be concentrated where patterns are more predictable, situations are simpler for algorithms to analyse, legal liability can be lower and fewer stakeholders are involved for interoperability.
Vendors and service providers should think about autonomous (and also remotely-driven) vehicles operating in areas such as:
- Mines and oil-fields (this is already happening, for example with autonomous trucks in big quarries)
- Farms, forests and agricultural areas (tractors, harvesters and similar)
- Construction sites (cranes, excavators and other similar units)
- Retirement villages (slower-moving golf carts and pods)
- Ports and airports
- Perhaps trucks in dedicated lanes on long-distance highways
- Autonomous boats and ferries on waterways.
This has some significant implications for network deployment and coverage, as many of these locations are in places with low population density, and with limited access to the general public. These are not sites that have, historically, been high priorities for most telecom operators. It will be interesting to observe whether this changes, or whether connectivity is more driven by private and niche providers than classical telcos.
The changing landscape of cities in the AV era
As discussed above, it is likely to be many years before we see traditional vehicles – especially private cars – displaced by fully-autonomous equivalents or new types of mobile “pod”. Nevertheless, it is worth thinking now about those eventual directions – and what that implies.
Potentially, this could change the nature of how we think about urbanisation, and where people tend to live/work. That in turn will affect how and where networks are built in future, and where connectivity is required.
Consider scenarios such as the end of large inner-city car-parks, if CAVs can be marshalled in remote locations, or shared more efficiently between multiple users. The nature of commuting may shift, allowing people to work efficiently both at home, and at “mobile desks” in CAVs of the future. Perhaps the density patterns of urban and rural areas will shift, as location becomes less essential – we already see ride-sharing enable people to go out in the evenings, without the need to worry about driving home or finding public transport. Wide adoption of CAV technology will likely intersect with human behaviour in many unpredictable ways.
This has not been the usual type of post about autonomous vehicles. Disruptive Analysis remains sceptical about many of the forecasts and timelines for full AV functionality to be both feasible, and widely permitted on the public road networks.
Telcos and their vendors need to be wary of the hype – but at the same time look for both near- and longer-term trends in the broader CAV space. There are opportunities that could be very significant – but it is important to take a holistic view, rather than being seduced by simplistic ideas about “selling a 5G slice” to an auto-maker for self-driving pods. There will need to be more infrastructure, and more focus on the intermediate goals and realizable use-cases first.
To learn more about ECI’s 5G Solutions, click here.