IAAI'06 Proceedings of the 18th conference on Innovative applications of artificial intelligence - Volume 2
T-drive: driving directions based on taxi trajectories
Proceedings of the 18th SIGSPATIAL International Conference on Advances in Geographic Information Systems
A personal route prediction system based on trajectory data mining
Information Sciences: an International Journal
Driving with knowledge from the physical world
Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining
MARiO: multi-attribute routing in open street map
SSTD'11 Proceedings of the 12th international conference on Advances in spatial and temporal databases
Computing with Spatial Trajectories
Computing with Spatial Trajectories
| Hi-index | 0.00 |
When analyzing the trajectories of cars, it often occurs that the selected route differs from the route a navigation system would propose. Thus, to predict routes actually being selected by real drivers, trajectory mining techniques predict routes based on observations rather than calculated paths. Most approaches to this task build statistical models for the likelihood that a user travels along certain segments of a road network. However, these models neglect the motivation of a user to prefer one route over its alternatives. Another shortcoming is that these models are only applicable if there is sufficient data for the given area and driver. In this paper, we propose a novel approach which models the motivation of a driver as a preference distribution in a multi-dimensional space of traversal costs, such as distance, traffic lights, left turns, congestion probability etc.. Given this preference distribution, it is possible to compute a shortest path which better reflects actual driving decisions. We propose an efficient algorithm for deriving a distribution function of the preference weightings of a user by comparing observed routes to a set of pareto-optimal paths. In our experiments, we show the efficiency of our new algorithm compared to a naive solution of the problem and derive example weighting distributions for real world trajectories.