Location based services are widely used in our daily life, especially in navigation, vehicle tracking and traffic flow analyzing. However, the GPS-equipped devices recorded trajectories are often inaccurate due to sensor failures, errors in detection techniques, disturbance signals, and different sampling rates. In order to reduce the inaccuracy of the collected trajectories, some existing works have already proposed, such as map-matching algorithm. Unfortunately, most current map-matching algorithms still have their own drawbacks and can be significantly improved.

In this study, we firstly evaluate state-of-the-art map-matching algorithms, then we propose a hybrid machine-crowdsourcing algorithm. In order to use the power of domain experts, we designed two types of crowdsourci...[ Read more ]

Location based services are widely used in our daily life, especially in navigation, vehicle tracking and traffic flow analyzing. However, the GPS-equipped devices recorded trajectories are often inaccurate due to sensor failures, errors in detection techniques, disturbance signals, and different sampling rates. In order to reduce the inaccuracy of the collected trajectories, some existing works have already proposed, such as map-matching algorithm. Unfortunately, most current map-matching algorithms still have their own drawbacks and can be significantly improved.

In this study, we firstly evaluate state-of-the-art map-matching algorithms, then we propose a hybrid machine-crowdsourcing algorithm. In order to use the power of domain experts, we designed two types of crowdsourcing questions: Direction Consult (DC) and Binary Direction Consult (BDC). By asking the experts these types of questions, we can let the experts to help us to reduce the uncertainty. Moreover, as a crowdsourcing-based algorithm, we make some efforts on managing the crowdsourcing questions, including crowdsourcing budget consuming and utilizing noisy crowdsourced answers. Last, we implemented our algorithm in both simulating and real world experiment, the result shows that the proposed method has better efficiency and effectiveness.