5G Will Not Be a Cure-All for Automated Vehicles

In a February 2019 episode of the Recode Media podcast, AT&T CEO Randall Stephenson began his case for 5G connectivity by citing automated vehicles (AVs). It’s not the first time he has done this, and he is not the only telecom executive to make this case. While continuing to improve wireless networks is a good thing, executives like Stephenson need to make a substantive case for the investment and AVs are not it. 

“Think about the applications that require instantaneous connectivity and the first one we all go to is autonomous cars,” Stephenson said. “You do not want to be in a car that’s relying on network connectivity if there is any latency.”

While that statement is certainly true, it is far from the most important aspect of connectivity for transportation. First and foremost is connectivity itself. If the vehicle cannot connect to a cell site in the first place, theoretical latency, speed, and network capacity have no relevance. 

Connectivity Remains an Issue

The San Francisco Bay area is a global hub of technology development, and even there connectivity can be an issue. Many of the world’s largest technology companies are based on the peninsula that stretches from San Jose to San Francisco. Yet when I was recently in the area, I only had to drive a few minutes north of the Golden Gate Bridge and turn off Highway 101 before I lost all connectivity entirely. 

Stephenson made the case that 5G could be used to replace the expensive and power-hungry compute platforms in AVs by doing all of the calculations from the sensors in the cloud and then sending control signals back to the vehicle. This is unlikely to happen anytime in the next several decades. 

AVs produce about 4 TB of data or more per day. Transmitting that much data to the cloud for processing seems a totally unrealistic expectation when there are more than 280 million vehicles in the US—tens of millions of which may be in use at any given moment. A single 1080p camera signal from an AV requires 3 Gbps and current Tesla vehicles with AutoPilot have eight cameras. 

The fastest 5G networks are expected to be those that use millimeter wave signals, but those have very short range and require about 10 times the number of micro-cell sites of current 4G networks for comparable coverage. Only about one in seven connections in the US are expected to be on 5G by 2025. Short range means more handoffs between cells which is also likely to problematic for real-time control of AVs. 

Improve Existing Network Coverage First

It’s not that connectivity isn’t essential for AVs. They need Wi-Fi and even LTE connections to enable software and map updates as well the ability to summon and dispatch the vehicles. I’ve written numerous times of the need for AVs to use vehicle-to-everything (V2X) communications in order to signal intent between each other and road users. However, V2X should not be essential to basic AV control since there may not always be anything to communicate with and the vehicle must be able to navigate its environment based on what the sensors can see. It is a means of extending situational awareness beyond the line of sight of sensors and cooperate in a complex driving environment. This capability can be done with either existing dedicated short-range communication or 4G C-V2X technologies. 

There are better reasons to justify 5G investments that telecomms executives need to emphasize while also investing in improving the coverage of existing networks.