Use multi-agent Reinforcement learning to dispatch several vehicles for comprehensive collection of urban information. We assume urban information is made up of small uniform grids, and every grid contain road networks and intersections. Vehicles can only move on road networks and intersections, and when vehicles arrive at a grid, this grid's information has been collected. In unit interval, every grid's information needs only to be collected once, no worth in going through a grid that has already been gone through. So optimization objective is maximizing the number of grids the vehicles passes over per unit interval.
- ITSC_project: this document saves open source code about experimental project published in the ITSC. Paper's name is "Multi-Agent Reinforcement Learning for Mobile Crowdsensing Systems with Dedicated Vehicles on Roads".
- new_reward_project: on-going experimental project.
See experimental_record for more detail.