Reinforcement Learning Based Advertising Strategy Using Crowdsensing Vehicular Data

As an effective tool, roadside digital billboard advertising is widely used to attract potential customers (e.g., drivers and passengers passing by the billboards) obtain commercial profit for advertiser, i.e., attracted customers' payment. The depends on number of customers, hence advertiser needs adopt strategy determine advertisement switching policy each as many possible. Whether a customer could be influenced numerous factors, such probability that see degree his/her interests in advertisement. Besides, cooperation competition among all billboards will also affect profit. Taking above factors into consideration, we formulate dynamic problem maximize advertiser. To address problem, first extract implicit information using vehicular data collected Mobile CrowdSensing (MCS), their trajectories preferences. With this information, then propose based multi-agent deep reinforcement learning. By proposed strategy, Extensive experiments three real-world datasets have been conducted verify our achieve superior compared with state-of-the-art strategies.