An on-line agglomerative clustering method for nonstationary data
Neural Computation
Approximation algorithms
The vehicle routing problem
Highly scalable trip grouping for large-scale collective transportation systems
EDBT '08 Proceedings of the 11th international conference on Extending database technology: Advances in database technology
NP-hardness of Euclidean sum-of-squares clustering
Machine Learning
| Hi-index | 0.00 |
In an effort to bridge the gap between public and personal transportation shared cabs are becoming attractive option, especially in large metropolitan areas. Places that see an aggregation of passengers such as airports and railway stations are well suited to serve as hubs from which vehicles are dispatched. While the standard vehicular routing problem has many solutions already, in this project, we consider two major variants of the problem - (a) The dynamic arrival of requests to destinations rather than a static set of such requests being used to solve the routing problem (b) The addition of a constraint that specifies the maximal acceptable deviation from the shortest route to a given destination. We model this problem as a variant of the On-line Capacitated Vehicle Routing Problem(CVRP) over the road network of a city where requests for cabs to specific destinations are viewed as data points in spatio-temporal space (the temporal aspect being the scheduling constraint before which the request MUST be fulfilled or rejected). We present a two phase solution approach. In the first phase, we cluster these data points to schedule the requests and reduce the problem size of CVRP. Then we solve each cluster as a CVRP. We seek to optimize the overall travel cost, maximize cab utilization and minimize waiting time of passengers. We have implemented the system and tested it with the road network data from Open Street Maps. Our experiments for the city of Mumbai (suburban area)shows that we can realize up to a 54% savings in travel cost with minimal infrastructure.