vehicle networ is what connects the different parts together as a vehile.
the lower-level basic
A few basic concepts: TCP/UDP socket, CAN Bus, Ethernet, MCU register, Ip address, MAC address, Ethernet driver, network configure, how to code a ECU/MCU etc.
there are EE students working on this close to physical level implementation, on how to talk with register, MCU pins and controls, and some implementing the data traffic protocol in network. While both are isolated from top applications, since no matter where the data come from, either AV application or connected vehicle or media data, doesn’t make a difference.
architecture
typical vehicle network architecture
beyond the basic, network architecture brings a good vision to understand what makes a vehicle electronicly as a whole. and which is serving all apps with v2x, in-vehicle entertainment, AV.
I was thought ECU handling events the same as web server. while they are very different. each ECU is doing only one special task. e.g. ABS ECU only dealing with ABS events, there are no many ABS events happening simultaneously to seize the ECU computing resource; however, a web server have to deal with multiplex request simultaneously, either using thread pool or asynchronous event callbacks.
each ECU actually has an easy life, but the bus seems easily choked since around 100+ ECU nodes in the vehicle network. however again, typical vehicle network architecture is very matured products, without data-heavy applications arising, current CAN bus is great.
smart phone in wheels?
one day in future, every vehicle running on road is requesting some data from cloud seamlessly, the scenario is like billions of browsers send requests to Google server seamlessly, then the cloud may face same issues in today’s web server.
Service oriented architecture(SOA) is also rising in vehicle network architeture design, but what kind of services is better locally in vehicle, and which services in cloud ? as hardwares evolution, even the computing-heavy tasks, e.g. vision based detection, path planning in AV, suppose not be a burden for local ECUs. so these services make sense locally served.
except that, I only image future vehicle as a smart phone on wheels. so remote cloud server do connect to, e.g. talk to another vehicle, play online game during driving, check weather, find out a parking lot, resturant or similar, or the car company want to steal user data privately?
SOA is a good move to decouple ECU modules and make the network bus light, and it’s fun to try some vehicle network architecture open project or play with ROS simulator if hardware required. but in a business view, I don’t know a good story to attract investors, namely vehicle network sounds not that blinking amazing.