Taxi journeys are usually priced according to the distance covered and time taken for the trip. Such a fixed cost strategy is simple to understand, but does not take into account the likelihood that a taxi can pick up additional pas- sengers at the original passenger’s destination. In this paper we investigate dynamic taxi pricing strategies. By using do- main knowledge, such strategies discount trips to locations containing many potential passengers, and increase fares to those areas with few potential passengers. Identifying a closed form optimal dynamic pricing strategy is difficult, and by rep- resenting the domain as an MDP, we can identify an optimal strategy for specific domains. We empirically compare such dynamic pricing strategies with fixed cost strategies, and sug- gest future extensions to this work.
|Number of pages||2|
|Publication status||Published - Aug 2014|
|Event||European Conference on Artificial Intelligence (ECAI-2014) - Prague, United Kingdom|
Duration: 18 Aug 2014 → 22 Aug 2014
|Conference||European Conference on Artificial Intelligence (ECAI-2014)|
|Period||18/08/14 → 22/08/14|