I have been very interested by the self driving car progress that has been made thus far. There are lot of interesting computer science problems to be solved to make self driving cars work. There are also lot of companies and some very big ones that are involved in making self driving cars a reality. Is there any niche space (like collecting more granular data) in this domain for a startup to exist?
It's going to be hard to out data collect Google (think of all the Android-phone miles and uses of Maps over the last few years on top of the Street View project), or even Tesla with it's field deployed systems. Even GM has years of OnStar data.<p>There might be a data play on dark-data...all those tables local engineering department traffic studies only accessible as PDF's, but it's hard to see that as non-replicable for a company that can stream data from a cars sensors to its servers in real time.<p>For a startup, there's an open question of how self-driving cars change the landscape of tier-two and tier-three automotive suppliers. What I mean is that what sort of new opportunities are made available by self-driving cars that are equivalent to manufacturing actuators or air-conditioning subassemblies.<p>The problem in the space is that the automotive industry is capital intensive. A billion dollars isn't going to ripple the pond and there are lots of companies with the habit of spending many times that. That's the way commodity industries work.<p>Good luck.
Think of new perspectives of the passengers and owners.<p>What would you do if you had a robotic chauffeur to take you wherever you wanted? Would you go places that you would not normally go to now? Cars drop you off at your destination and park themselves (maybe call you to agree to whatever the lot or garage, recharge station is charging)<p>SD-Cars make the driver the passenger... who will now be able to look out and see the roses... listen to an interactive on-the-raod tour. Read the billboards, notice that fruit stand etc. Former drivers will be asking "are we there yet?" Parents can truly interact with their kids on the road.<p>People that didn't travel as much due to health or medical reasons may be traveling more.<p>A lot more day planners, maps, review sites, etc. will probably interact more with the vehicle...
I've been working on some for the past years. You should think of the interactions a vehicle will have. There are:<p>- vehicle to vehicle (V2V)<p>- vehicle to infrastructure (V2I)<p>- vehicle to cloud (V2C)<p>Each of those niches has niches of their own. I just finished working on the V2C niche by developing an API to move data between vehicles and updating them automatically (not realtime, though). What can you do with that kind of data? Well, if you have cars operating data (engine temperature, speed, etc.) you can come up with a bunch of interesting stuff.<p>Don't wait too much. This self driving market is reaching peak hype soon.
Data. A kaggle for self driving data would be fantastic.<p>As a computer vision researcher, getting my hands on data 'at scale' is a really tough problem.