In grad school I did research on a on increasing the average vehicle occupancy in cities and we took a lot of inspiration from the bike sharing programs.
Have you considered offering dynamic pricing based on the likelihood that the bike will be used in the near future? I.e. riding into the city should be cheaper than riding out of the city in most cases.
Edit: a few of us did graduate transportation research as well, your work sounds cool. Is it published anywhere?
[1] http://www.cdtm.de/events/entrepreneurship-camp/about-cdtm/b...
a) a heuristic for optimal routing of repositioning trucks based on anticipated (uncertain) future demand.
b) dynamic pricing using a model predictive control approach (receding horizon optimization on a model of customer reaction to (price) incentives).
Insanely cool stuff. And its working. A paper should come out soon..