On Map Matching of Wireless Positioning Data: A Selective Look-ahead Approach Matt Weber 1,* , Ling Liu 1 , Kipp Jones 2 , Michael J. Covington 1 , Lama Nachman 3 , and Peter Pesti 1 1 Georgia Institute of Technology 2 Skyhook Wireless 3 Intel Corporation ABSTRACT Wireless Positioning Systems (WPS) are popular alterna- tive localization methods, especially in dense urban areas where GPS has known limitations. Map-matching (MM) has been used as an approach to improve the accuracy of the estimated locations of WiFi Access Points (APs), and thus the accuracy of a wireless positioning system. Large-scale wireless positioning differs from satellite based positioning in at least two aspects: First, wireless positioning systems typically derive the location estimates based on war-driving access point (AP) data. Second, the locations of the AP beacons are not generally known at the same precision as that of the satellite locations. This results in lower accu- racy and a lower confidence factor in the use of wireless positioning. This paper presents a fast selective look-ahead map-matching technique, called SLAMM. Existing MM al- gorithms developed for real-time location tracking of a mov- ing vehicle are ill-suited for matching large collections of war- driving data due to the time complexity. Another unique feature of SLAMM is the map-matching of critical location samples in an AP trace to the road network before matching non-critical samples. Our experiments over a real dataset of 70 million AP samples show that SLAMM is accurate and significantly faster than the traditional MM approaches. 1. INTRODUCTION Large-scale WPS typically determine the position of a WiFi enabled mobile client based on the geographical loca- tions of its observed wireless APs. War-driving is a popular means of mapping observed WiFi APs to locations via GPS. For example, the WPS solution provided by Skyhook Wire- less[1] delivers location estimates based on a user’s surround- ing wireless APs and cellular towers. Skyhook Wireless em- ploys numerous drivers in different regions of the world to ∗ To whom correspondence should be addressed. Email: mattweb@ gatech.edu Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Copyright 20XX ACM X-XXXXX-XX-X/XX/XX ...$10.00. drive around and record the signal strengths of observed APs and cellular towers along with the corresponding GPS read- ings, a process known as war-driving. The result is roughly 1,400 hours of GPS data uploaded daily. MM refers to a process of matching location samples to a road network. In this paper, we focus on matching very large datasets of trajectories with location samples collected at once per second. We argue that MM can assist in both the accuracy of the final location estimation and the manage- ment of the war-driving data collection process. However, to achieve these enhancements, both MM accuracy and speed are required. Thus, the unique challenge for matching large AP datasets is twofold: First, we need a high quality MM algorithm that can utilize available global knowledge to in- crease the accuracy of AP location data, while leveraging the speed of incremental matching algorithms.Second, the MM approach should make use of the road network topology to enhance the accuracy of incremental techniques, and must be resilient to both errors in the location measurement and error in the road network. With these design objectives in mind, we develop a se- lective look-ahead MM approach, called SLAMM. SLAMM employs three filtering techniques that step by step identify and match critical location samples before matching the rest. In addition, the matched samples are given a road segment based quality estimate unlike any in current literature. We compare the speed and accuracy of SLAMM to a tra- ditional incremental MM approach from a real AP dataset of 70 million location samples, and show that the SLAMM three-level progressive filtering based MM approach is accu- rate and significantly faster than traditional approaches. The remainder of the paper is structured as follows. In section 2 we discuss the unique challenges with matching commercial war-driving data. In section 3 we provide an overview of existing work and review two basic MM ap- proaches. In section 4 we present SLAMM in detail. In section 5 we evaluate SLAMM by comparing it to two basic MM approaches. In section 6 we discuss related work, and finally in section 7 we give our conclusion and discussion for future work. 2. WAR-DRIVING CHALLENGES The study conducted in this paper uses the war-driving dataset collected by a fleet of drivers for Skyhook Wireless, who systematically drive in tens of thousands of cities and towns worldwide to scan for 802.11 WiFi APs. The data