Citi Bike redistribution optimization

Problem

Uncertain arrivals, departures, and station capacity can leave riders facing empty stations or full docks. I investigated the tradeoffs in redistribution strategies.

Approach How I thought about it

• I modeled finite-capacity stations with continuous-time Markov chains, birth-death processes, and queueing methods.
• I compared stochastic demand and truck redistribution through steady-state system performance.

What I built or did

• I developed a stochastic bicycle availability and redistribution model for five Citi Bike stations in Midtown Manhattan.

Results

• Modeled result

  Availability failures decreased from approximately 41.87 per hour to 1.77 per hour in a scenario using five truck visits per hour. This is model performance, not observed Citi Bike operations.

Tools

• Continuous-time Markov chains
• Queueing theory
• Stochastic modeling
• Steady-state analysis