EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1025. Spatio-temporal pricing of battery-swapping tasks on mobility-sharing platforms using proximal policy optimization
Invited abstract in session TC-43: Simulation in transportation and logistics, stream Agent-based Models in Management, Economic and Organisation Sciences.
Tuesday, 12:30-14:00Room: 99 (building: 306)
Authors (first author is the speaker)
1. | Minjeong Kim
|
2. | ILKYEONG MOON
|
Industrial Engineering, Seoul National University |
Abstract
Battery-powered mobility-sharing companies face operational challenges in managing low-battery mobilities due to the dockless service, allowing pickups and drop-offs anywhere. To tackle this, such companies engage independent contractors to recharge these mobilities, offering them per-task compensation. Given the contractors’ liberty to dictate their work schedules and volumes, it is essential to devise a pricing strategy to motivate their participation. The core tradeoff involves balancing the wages paid to workers against the penalty costs from unsatisfied customers due to incomplete charging tasks. To this end, this study proposes a spatio-temporal pricing strategy that assigns tasks with differentiated regional prices by time interval, aiming to attract workers to areas with a scarcity of available workers. We use a reinforcement learning approach with proximal policy optimization to handle the high dimensionality of the problem. A simulation environment mimics dynamic worker participation and task reservation in response to price updates made by the RL agent. Computational experiments demonstrate the effectiveness of reducing overall platform costs under varying task and worker distributions. Moreover, the findings indicate that the RL-driven pricing policy mitigates the spatial and temporal discrepancies between task demand and worker availability. Our study is applicable to other spatial crowdsourcing platforms which necessitate resolving spatio-temporal imbalances.
Keywords
- Artificial Intelligence
- Transportation
- Computer Science/Applications
Status: accepted
Back to the list of papers