Hardware Costs for Self-Driving Cars

The news over the past several months has been buzzing with discussions of the pending development and deployment of self-driving cars.  At the forefront of this discussion has been Google, who has been working on this capability since around 2009 (Google, n.d.). 

A lot of the discussion in regards to the advancement of this capability has centered on the software development and extensive testing that is being put into maturing this capability.  But once this non-recurring engineering has been completed, or at least matured to some level that incremental updates will not be as significant, it’s assumed that consumers will now be mainly paying for the hardware that supports this technology.  But in today’s dollars, how much does that hardware actually cost?  That’s the question that Alex Davies from Wired Magazine set out to identify. 

In Davies’ article, he cites a report by the Boston Consulting Group as forecasting that “you can have a car that parks itself for an extra $2,000” but a fully autonomous car will “add $10,000 to the pricetag” (Mosquet, 2015).  The article goes on to break down the hardware associated with this capability into three main categories: “sensors, processors, and actuators.” 

It can be argued that the actuators and computer processors needed to support this capability are fairly mature and even deployed currently in a lot of cars on the road today.  I agree with the article’s position that there is “room for more speed and the costs to come down a bit”, which is good news for consumers. 

So from the sensors perspective, the article lists GPS, Lidar, cameras (it’s assumed that they’re talking both infrared and electro-optical), ultrasonic, odometry, and Radar systems.  The two big-cost systems that the article mentions are the GPS (estimated to cost from $80 to $6,000) and Lidar ($90 to $8,000).  For the cost of the GPS, this is for a high end system that is using some type of differential GPS that is coupled with other sensors such as an altimeter and gyros to enhance the accuracy.  The article doesn’t mention this, but it’s also assumed that there will be some flavor of Kalman filtering going on within the system in order to protect the system for spurious or erroneous inputs effecting the overall output from the system.  The other sensors mentioned in the article all range from around $15 to $200, which I feel in the overall cost of the system, is insignificant.

     

The key takeaways for me from this article are that:

-          There is work to be done to bring down costs of position and Lidar sensors

-          Adding additional sensors to provide greater coverage and/or redundancy should be pursued

-         There are multiple sensor options available that provide a diverse amount of input data to the system in order to provide the greatest possible picture of the operating world to the vehicle

References:

• Davies, A. (2015, April 4). Turns Out the Hardware in Self-Driving Cars Is Pretty Cheap. Retrieved October 25, 2015, from http://www.wired.com/2015/04/cost-of-sensors-autonomous-cars/
• Google Self-Driving Car Project. (n.d.). Retrieved October 25, 2015, from https://www.google.com/selfdrivingcar/where/
• Mosquet, X. (2015, April 21). Revolution in the Driver's Seat: The Road to Autonomous Vehicles. Retrieved October 25, 2015, from https://www.bcgperspectives.com/content/articles/automotive-consumer-insight-revolution-drivers-seat-road-autonomous-vehicles/