Dynamic Public Resource Allocation Based on Human Mobility Prediction
summary
Formulates public-resource movement as a multi-agent long-term maximal coverage scheduling problem and combines multi-step crowd flow prediction with Model Predictive Control to replan during the day. Implements a two-step heuristic Energy Adaptive Scheduling (EADS): a multi-level max-cover phase and an energy-aware hill-climbing phase, evaluated on Beijing Happy Valley crowd flows and actual trash-bin locations. EADS cuts required resources by 80.0% versus current deployment and 56.8% versus static maximal-coverage, while respecting agent energy and start/return constraints.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:17Background
- 0:38Motivation
- 1:33Opportunity
- 1:52Challenges & Solutions
- 3:11Multi-agent Long-term Maximal Coverage Scheduling
- 4:40Energy Adaptive Scheduling
- 5:26Real-World Datasets
- 5:37Evaluation
related talks
01h 32mOperating Bike-Sharing Systems – A Case Study in Real-Time Decision Making
01h 10mCrowd Your Way Out of a Paper Bag
06mMulti-Head Spatio-Temporal Attention Mechanism For Urban Anomaly Event Prediction
01h 03mUnderstanding and Shaping Urban Mobility Using Dollars and Data II
29mIncentives and Games in Large Scale Mobility
32mLean Experimentation and Decision Science with Humans in Warehouses
06mPassive Health Monitoring Using Large Scale Mobility Data
12mReducing Carbon Footprint in Smart Buildings with Data Science