Smarter Tools for (Citi)Bike-Sharing
summary
City bike data (≈7,000 bikes, 14.5k docks, ≈400 stations, 1.2M trips/month) drives a model that minimizes expected unhappy users (arrive-to-rent with no bike or arrive-to-return with no dock). Tools: station-level continuous-time Markov chains, time-varying Poisson rates, simulation-optimization and an LP-relaxation solved by a greedy polynomial algorithm that provably optimizes joint bike/dock allocation. Results: optimal reallocation with overnight rebalancing yields ≈9,300 expected upset customers vs current layouts at ≈12,000; simulation gives modest extra gains (~3–20% improvements depending on model). Operational constraints and politics limit deployment.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:33What is a bike sharing system
- 2:11A day in the life
- 6:12Data
- 10:25Target Levels
- 11:03Basic Framework
- 13:04Markov Chain
- 13:51Planning Cycle
- 15:27System Problem
- 16:34Data Analysis
- 23:50Data Visualization
- 50:27Incentives
related talks
06mUnderstanding User Behavior in Car Sharing Services Through The Lens of Mobility: Mixing ...
39mApidays Virtual Helsinki - Hendrik Wolff, Simon Fraser University
06mData-Driven Battery-Lifetime-Aware Scheduling for Electric Bus Fleets
28mContainer Tracking for Bosch Tools Supply Chain Operations
16mTidy Transit: Real Life Data Modeling for Public Transportation
06mRISC: Resource-Constrained Urban Sensing Task Scheduling Based on Commercial Fleets
17mA comparison of SUMO’s count based and countless demand generation tools
21m"Capacity Prediction" instead of "Capacity Planning":