DeepPool: Distributed Model-Free Algorithm for Ride-Sharing Using Deep Reinforcement Learning

Abstract

The success of modern ride-sharing platforms crucially depends on the profit of the ride-sharing fleet operating companies, and how efficiently the resources are managed. Further, ride-sharing allows sharing costs and, hence, reduces the congestion and emission by making better use of vehicle capacities. In this work, we develop a distributed model-free, DeepPool, that uses deep Q-network (DQN) techniques to learn optimal dispatch policies by interacting with the environment. Further, DeepPool efficiently incorporates travel demand statistics and deep learning models to manage dispatching vehicles for improved ride sharing services. Using real-world dataset of taxi trip records in New York City, DeepPool performs better than other strategies, proposed in the literature, that do not consider ride sharing or do not dispatch the vehicles to regions where the future demand is anticipated. Finally, DeepPool can adapt rapidly to dynamic environments since it is implemented in a distributed manner in which each vehicle solves its own DQN individually without coordination.

A Probabilistic Approach for Demand-Aware Ride-Sharing Optimization

“…Second, upon change in vehicle status, e.g., a user is delivered to the destination or an empty car picks up a new user, the system can re-optimize the routing decisions and request-vehicle assignments, so as to optimize the long-term ride-sharing performance in a greedy fashion. The overall solution can serve as a baseline for other demand-aware studies, e.g., the conceivable ones by extending the approach in [13] and [27] to the multi-rider setting and the recent one in [5]. We leave the performance analysis of the overall greedy solution, as well as developing solutions with optimized long-term performance, as interesting and important future directions.…”

“…We assign a representative node in each region, from which all other nodes in the region can be reached in a small number of slots, e.g., 5 slots, illustrated as the black dots in Fig. 1(b) 5 . We add a directed edge (u, v) in G, as illustrated in 1(a), if there exists a path from the representative node of region u to that of region v in the road graph such that the major 5 For a general urban road network, constructing the regions is equivalent to solving a clustering problem to find a set of clusters.…”

“…1(b) 5 . We add a directed edge (u, v) in G, as illustrated in 1(a), if there exists a path from the representative node of region u to that of region v in the road graph such that the major 5 For a general urban road network, constructing the regions is equivalent to solving a clustering problem to find a set of clusters. Location points within each cluster are close to each other.…”

“…should not be larger than p t v,u,k , the probability that the k requests appear. The inequalities in (5) state that the assignment probability can be positive only if the assignment is feasible (see Definition 1). The inequalities in (6) can be understood as follows.…”

Artificial Intelligence driven optimal route planning for urban transit

“…In (Al-Abbasi et al, 2019), the authors obtained a real-world dataset of taxi trips in Manhattan, New York City, they used the data for June 2016 for training neural networks and used one week's data form July 2017 for evaluation purposes, the dataset contained numerous attributes however for the research they made use of the pick-up time, location, passenger count and drop-off locations to develop a travel demand predictive model.…”