Key technologies for autonomous vehicles (4/6)

3 min read

If self-driving vehicles are to hit public roads, and deliver the plethora of potential benefits promised, the communication aspect needs to be fully addressed.

Onboard sensors can only provide some of the information about a car’s environment. It is clear that a more expansive map of the surroundings will be needed to help the car decide on the correct course of action. This will be created using information from other road users and infrastructure in the vicinity.

Vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication will both rely on the use of wireless or cellular systems, but which protocols are the best fit remains to be determined. The winners here need to offer high data transfer rates (both upstream and downstream), plus low-latency operation to deliver rapid responsiveness and thereby improve safety. In addition, high degrees of security will be mandated – so that vehicles can’t be hacked by malicious third parties.

Car manufacturers, tier one suppliers, silicon vendors and telecoms companies are all currently collaborating on the development of various forms of V2V/V2I technology to incorporate into the next generation of vehicle models. Let’s look at the main candidates that have emerged.

Dedicated short-range communications (DSRC) is arguably the best-known inter-vehicle communication method to date. This has been employed within the automotive space since 1999. Based on the 802.11p wireless communication protocol, DSRC runs between the 5.8GHz and 5.9GHz bands (depending on the country implementing it) with a range of up to 300m. Its relatively low latency rate of roughly 5ms means it offers huge benefits in terms of safety compared to other communications technologies.

Many companies have looked at DSRC, including Toyota, General Motors and Ford, as a cost-effective method to achieve their communications goals. However, although it is widely understood and inexpensive it still has a number of drawbacks, and these are restricting its attraction. As DSRC bandwidths differ around the world, it is more challenging to integrate it for the global market, where cars are almost identical whether sold in Europe, Asia, North America or elsewhere. Security is also an issue. Data would likely be encrypted, but DSRC has been proven to be vulnerable to jamming, false alerts and so called “man-in-the-middle” attacks (where hackers alter the communications between two parties). Finally, it only supports small data rates (measured in Mbps), when prototype autonomous vehicles are already at Gbps levels, just from the information acquired by their sensing mechanisms (such as LiDAR imaging).

The limited future appeal of DSRC is a position proven, in part, by Ford’s decision to focus on cellular networks. Many automotive firms have already skipped LTE 4G technology completely though, because although infrastructure is already in place, it doesn’t provide the necessary performance to support autonomous vehicles. LTE networks suffer from serious operational constraints, with peak data rates of only 300Mbps and an average latency of around 50ms. Instead the majority of companies are looking to next generation 5G networks as the technology to form the foundation of V2V and V2X systems.

Special-interest groups, such as the 5G Automotive Association, are bringing car manufacturers, suppliers, chip makers and telecom network operators together to develop the technology for automotive applications. 5G’s benefits are obvious, delivering incredibly high data rates that could go well beyond 10Gbps, exceptionally low latency of 1ms, plus the ability to leverage existing infrastructure.

Speed is key. If you downloaded 4.5GB of data (a film, for example) over a household DSL line, with a typical speed of 50Mbps, it would take 13 minutes. Over a 5G connection it could be done in as little as 4s. Even with the amount of information required for self-driving vehicles, 5G means data will be able to flow freely, without the need for buffering or prioritizing certain packets of information over others. But probably the biggest advantage 5G brings to the table is its ultra-low latency, a key element in autonomous operations.

The time it takes a human to recognize and then react to an incident on the road is about 1s. If they are in a car traveling at 100km/h, some 28m will be covered before they hit the brakes and begin to slow down.

A connected vehicle, with a high-speed 5G connection, could react 1,000 times faster, beginning a braking procedure after traveling just a few cm.

It must be accepted that there are still serious hurdles to overcome, though – particularly in relation to security. The UK government has pinpointed four areas in need of attention.

  • Cross-layer security: Where a unified framework is needed to coordinate different security methods for each security layer.
  • Cross-domain security: 5G networks create a huge amount of different use cases, so cooperation between groups will be needed to integrate security solutions across domains.
  • End-to-end security: There should be a secure connection for the communication paths between the user and the core network.
  • Security-by-design: As the network evolves, security must be built into the design during development.

Perhaps more importantly than this, network coverage needs to expand beyond what we are currently familiar with if connected autonomous vehicles are to be adopted by the mainstream. When you’re walking around rural areas with your mobile and the cellular network drops out, it is frustrating – but far from critical. Conversely, if a connected autonomous vehicle, relying on external information, were to lose its signal as it approached a busy or complicated section of the road network, it could be disastrous.

Communication networks of the future will not only have to be incredibly robust, with huge amounts of redundancy, but also expand to include a greater geographical area – reaching communities that currently aren’t even well served by 4G technology. Until that stage vehicle autonomy is only likely to be applicable in metropolitan areas.