0018-9545 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TVT.2016.2639550, IEEE Transactions on Vehicular Technology 1 MA-SSR: A Memetic Algorithm for Skyline Scenic Routes Planning Leveraging Heterogeneous User-generated Digital Footprints Chao Chen, Xia Chen, Leye Wang, Xiaojuan Ma, Zhu Wang, Kai Liu, Bin Guo and Zhen Zhou Abstract—Most of the existing trip planning work ignores the issue of planning the detailed travel routes between Point of Interests (POIs), leaving the task to online map services or commercial GPS navigators. However, such a service or navigator cannot meet the diverse requirements of users. Particularity, in this paper, we aim at planning travel routes that not only minimize the distance but also provide high-quality sceneries along. To this end, we propose a novel two-phase framework to plan travel routes efficiently in a large road network considering multiple criteria, i.e., both the quality of the scenic view and the travel distance. In the first phase, we enrich the edges and assign a proper scenic view score for each of them by extracting relevant information from heterogeneous digital footprints of geo-tagged images and check-ins. In the second phase, on the top of the enriched road network, given users’ trip queries, we employ the concept of skyline operator and propose a memetic algorithm (MA) to discover a set of equally optimal routes with diverse travel distances for users to pick from. Finally, to validate the ef- ficiency and effectiveness of the proposed framework, we conduct extensive experiments on three real-world data sets from the Bay Area in the city of San Francisco, which contain a road network with 3,771 nodes and 5,940 edges crawled from OpenStreetMap, more than 31,000 geo-tagged images generated by 1,571 Flickr users in one year, and 110,214 check-ins left by 15,680 Foursquare users in six months. Results demonstrate that MA yields high- quality solutions that are reasonably close to the optima but within desirable computation time, and considerably better than the baseline solutions obtained by Genetic Algorithms. Index Terms—Scenic View; Skyline Routes; Memetic Algo- rithm; Heterogeneous; Digital Footprints I. I NTRODUCTION P Lanning an itinerary before paying a visit to a city is one of the most important preparation activities [8]. To Copyright ©2015 IEEE. Personal use of this material is permitted. However, permission to use this material for any other purposes must be obtained from the IEEE by sending a request to pubs-permissions@ieee.org. Chao Chen and Kai Liu are with the Key Laboratory of Dependable Service Computing in Cyber Physical Society (Chongqing University), Ministry of Education and also with the College of Computer Science, Chongqing Uni- versity, Chongqing, China. E-mail: {ivanchao.chen, liukai0807}@gmail.com. Xia Chen is with Chongqing Automotive Collaborative Innovation Center, Chongqing University, Chongqing, China. Email: chenxia_office@126.com. Leye Wang and Xiaojuan Ma are with the Department of Computer Science and Engineering, Hong Kong University of Science and Tech- nology, Clear Water Bay, Kowloon, Hong Kong. Email: {wangleye, xiao- juan.c.ma}@gmail.com. Zhu Wang and Bin Guo are with the Department of Computer Sci- ence, Northwestern Polytechnical University, Xi’an, China. E-mail: {transit- wang,guobin.keio}@gmail.com. Zhen Zhou is with the School of Management, Northwestern Polytechnical University, Xi’an, China. E-mail: zhouzhen@nwpu.edu.cn. figure out a satisfactory travel route, a user may not only have to browse as many profiles of Point of Interests (POIs) as possible to pick up his/her preferred one(s), but also need to determine the order of travelling, which is known to be very time-consuming and labor-intensive [10], [38], [46]. To ease the process of travel planning, a few of online trip planners have been developed to rank popular city landmarks to guide users to select the interesting places, and also help to organize their travelling order as well [7]. However, before composing a travel route, additional time and efforts from users to take part in iterations with machines are needed (i.e., not automat- ically). Fortunately, with the increasing popularity of location- based social networks (LBSNs), various user-generated digital footprints that record the interactions between human and the cyber-physical world are accumulated at an unprecedented scale and speed [16], [27], [47]. These digital footprints have rich and valuable information regarding the POIs and users, such as the physical coordinate, category, popularity of the POI embedded explicitly or implicitly, making completely automatic trip planning under different user scenarios possible through mining such data [25], [50]. However, most current automatic trip planning systems rec- ommend either a single POI or a sequence of POIs, neglecting the detailed travel route planning issue between two suggested consecutive POIs. Although available on-line map services or commercial GPS navigators can be easily integrated and provide travel routes in terms of the shortest travel distance or time, such a service cannot meet the diverse user requirements in many cases [30], [44]. As an instance, in the case of travel by driving for leisure purpose, the scenic view along the travel routes would be of great importance to users, and a good planning service should also put the sceneries along the route in a high priority [4], [5]. Thus, beyond the shortest distance (or time) taken, in the paper, we also take the quality of scenic view along the suggested routes into account, with the objective of planning travel routes that not only minimize distance but also offer high-quality of sceneries along. Intuitively, to plan an optimal travel route from one POI to another one considering both the travel distance (or time) and quality of the scenery criteria, for each edge in the road network, we can first assign a score for each factor (i.e., the scenic view score and the distance respectively) and define a weighting score function to aggregate them into a combined score according to the user preference. Then, the travel route composing by edges with the highest total combined score under a user-specified budget (e.g., the travel