Honestly, there is no way from "here to there" when it comes to fuel efficiency from an ICE. We are hovering around ~30% efficiency for modern mid-sized automobiles. Some estimates put that at a lower figure. The maximum efficiency theoretically possible is limited by the Carnot cycle, and I think it's ~60% if IIRC. There are two other factors you can play with: weight and energy recovery.
As far as weight goes, heavier cars are actually MORE efficient (weight to fuel wise). It's why buses are more efficient than cars. Believe it or not, a tractor trailer getting ~4-6mpg is way more efficient than a Honda Accord. It's carrying 80,000 lbs and the Honda is only moving about 3000 lbs. This argument doesn't hold much water though when you simply talking about people moving. The tendency in the US is for everyone to drive their own car. Therefore, the person-miles/gallon is fairly low but this is really about weight efficiency. If I move a 200 lbs object with a 3000 lbs one, my weight efficiency ratio is less than 1:10. Adding people just raises that ratio. The other option is to lower the weight of the transportation.
This is tough to do, and keep cars safe. Most increases in automobile safety has come from: collapsible steering wheels, seat belts, and crumple zones. Don't expect that other "industrial" vehicles will go down in weight though. They may make the vehicle lighter, but the load will just go up. It will still be 80,000 lbs tractor trailers vs 3000 lbs vehicles. There is a point at which no amount of crumple zones will save you when these two things collide. A fix for this side effect might be self driving cars that nearly never crash. Though, in this scenario you make crashes less likely, but increase their rate of fatality.
As for energy recovery it seems that the mechanical/electrical cycle provided by batteries is one of the best, but don't expect it to improve highway figures by much. Around town there still could be some improvement, as wind resistance is low as so is friction. The highway is a different matter, and that is evidenced by the current figures from existing hybrids. The only way to improve those numbers is to reduce friction and wind resistance. One is materials science (friction) and I'm sure it's possible but pricey. Options there must be carefully weighed to ensure that what ever new near friction-less material is sustainable and doesn't cause more CO2 just to make it compared to the fuel savings. A second option (wind resistance) is largely based on aesthetics. Will people buy cars that look funny? Hard to answer that one as tastes change.