Grant: Alliance Grants

The growing trend of auto dependency has led to a shortage of parking spaces in major urban cities. The mismatch in parking supply and demand has adverse consequences such as long parking search times and traffic congestion. Dynamic pricing is a parking management strategy that adjusts parking rates in real time according to the monitored parking occupancy levels obtained from sensors. When occupancy levels are high the rates are increased to open more spaces, and when occupancy is low, the rates are decreased to attract more drivers. The main challenge in the implementation of dynamic pricing is to deploy sensors that are scalable, efficient, and cost-effective, for capturing real-time parking occupancies to support the price-setting mechanism. Parking sensors, however, have not been successful in the long term due to their high acquisition costs, inefficient scalability, and prediction inaccuracy. The objective of this project is to use parking payment data instead of sensors for the purpose of predicting real-time parking occupancy, which can be used to derive optimal parking rates in dynamic pricing. The transaction data-dependent occupancy predictions can also aid policymakers in implementing optimal parking regulations and pricing strategies for the purpose of improving the overall state of the transportation network. This project proposes a data-driven methodology using Machine Learning and Stochastic Optimization for predicting parking occupancy levels from parking transaction data and adjusting the rates dynamically to balance the supply and demand of parking in an equitable fashion that maximizes the social well-being of all road users.