In my master thesis I want to compare 2 diff. computation approaches (most likely - cloud and edge) in order to look, which approach is best suited for which use cases for a connected car.
For example, "edge technology is suited to autonomous parking, because ... latency etc..."
or
"cloud is good for predictive maintenance, because..."
I know how I want to execute the analysis:
1. Define different disciplines of autonomous driving
2. Compare technical performance (based on diff. attributes like latency, data volume etc.) of diff. approaches (like cloud and edge)
3. Make a matrix where a technological approaches matches a "situation".
But I don't know how to call this method. And maybe there are some recommendations / best practice cases of how to do that.
Thank you in advance!
The first thing to investigate is KPIs, key performance indicators for solving the problems you list, specifically. Then you can compare and benchmark.
You could also review past literature about distributed computing or centralised computing. You will find out that cyclically, computing increases distribution (de-centralising) during a certain phase (now it's the move to the Edge), and then comes back to a top-down centralised approach. Let me give you several examples in the long history of computing: the abacus is a personal computer used in Ancient China, hence this is distributed computing; so is the PC (personal computer), however the Supercomputers like the Cray machines are centralised computers to which remote users submit computation tasks called batches.
Therefore, structurally, there is nothing fundamentally new in centralisation vs distribution. Already in the 1980s, the benefits of each were clear to computer scientists, and you will find such comparisons.
This goes even into the hardware design of a computer: do you go for a main single chip which is very powerful, or do you put on a board several chips less powerful each but combining their strength efficiently? A GPU, a DSP, a CPU, all these appeared when the idea seemed best, but things change.
Can I also suggest you benchmark the computational approach of the USA and of the USSR from 1960 to 1980? Very schematically you will find a very deep effort in cracking the theoretical nut first (math) and using simpler computers to finish the resolution of a problem in USSR, and a very limited theoretical effort initially and straight massive computing on the most sophisticated computers in the USA.
This is of course a gut-feeling, which needs further assessment, but please bear this in mind, while developing your own view and set of KPIs.
Systems can be designed for robustness and longevity, with no shortcoming, or they can be designed for power and features, taking some chances.
Neither one nor the other is always best...
I hope this helps
Dear Renaud Di Francesco , thank you for your elaborate response. It sound like a good academic approach and you gave some nice ideas, regarding "evolutionary" perspective and comparison of developments in different countries. I've just recommended your answer, since it is good recommendation from a researcher affiliated with IEEE.
- Badges
- Science topic
- Similar topics
- Computer Science