Hi Kris. The US government requires seat belts, air bags and antilock brakes. They just added a requirement for backup cameras starting in 2018. The governement is very involved in requiring new features in cars for our safety.
IMO USA car companies see electronics as a way to boost profits by adding high-margin features to what is otherwise a low-margin product. What they fail to realize is that as new cars get more expensive, fewer people can afford them. Indeed, young people with student loans to pay back won't be in the market for a car for years or decades after graduation, and have discovered that living and working in an urban area with reasonable public transportation is a better life than car-oriented suburbs. On the few occasions when they need to bring home something heavy, there's car-share.
It's ironic that the Microchip thread is talking about how technology has made ICs so cheap that nobody can make money from them, while cars -- which have been around for well over 100 years -- keep getting more and more expensive. Maybe the IC makers need to convince consumers (or government safety agencies) that all ICs need air bags or something :-)
For self parking you need computer vision algorithms that get confused with compressed data. Also time lag is an issue (which is partially corrected with Ethernet timestamping).
For park assistance, there is no computer vision algorithms involved. Only video enhancement, fish eye correction and different kinds of videos stitching (2D and/or 3D mapping). Those routines work well with compressed videos.
We don't want to use shielded cables because they are expensive and less robust (temperature, vibration, wear) than twisted pair. That's why BroadR Reach was created: only one pair for bidirectional Ethernet, controled signal rise/fall time for less radiated emission and increased robustness to ESD and induced perturbations. Same wire pair as for CAN bus.
BroadR Reach is new, in the meantime we use LVDS for uncompressed video with shielded cables. APIX is an enhanced LVDS link for high datarates in an automotive environment (with modulated side channels for I2C-like communication).
I would need to be convinced that even a self-parking system would require uncompressed video. The objects that need to be detected for this self-parking feature, even down to the size of a pebble, can easily be resolved with motion JPEG or MPEG, at data rates that are way lower than what uncompressed video HD would require. I think perhaps the biggest concern in a self-parking systems would be to keep the lag created by compression as low as required. That can be tuned in the algorithm.
As to cabling, I'm actually surprised that automotive Ethernet wouldn't use shielded cat-5e, as opposed to unshielded, although clearly, unshielded is preferable because it's cheaper. All depends on what the EMI requirements are.
I appreciate that car vendors seek to upsell their customers to the ultimate package of bells and whistles. Audio equipment components in a car may cost 10x what they cost as consumer goods at home. That said, I'm wondering what the electronic component costs are for park assist and to what degree hardware cost reductions make any meaningful difference. Aren't most of the costs of these features the car company's markup and mandatory bundling of other features to sell the total premium package?