A City-Wide Crowdsourcing Delivery System with Reinforcement Learning
summary
A crowdsourced, public-transport delivery system uses DQN with time-estimation filtering to pick next-hop stations for packages riding subways and buses; training uses three months of transit traces plus one month of on-demand package logs and 1 million orders from Shenzhen. The method models time constraints, multi-hop transfers, and profit, yielding a 40% higher profit rate and 29% higher on-time delivery versus baselines and an extra 9% profit and 8% delivery gain from action filtering; data is shared.
AI-written summary · about
related talks
35mCrowd dynamics, RL and Unity: A Journey
11mUsing Deep Reinforcement Learning to Coordinate Multi-Modal Journey Planning with Limited ...
19mTestbed for Development of AI Solutions to Boost Operational Efficiency in Transportation Networks
21mChallenges in Reward Design for Reinforcement Learning-based Traffic Signal Control
22mMicrodosing the Economy
18mExtending Mobility - How AI Contributes towards the Future of Railway Transportation
28m"Where, When, Why, and How: Directi
14mHuman-in-the-Loop Reinforcement Learning