http://ito-lab.naist.jp/themes/pdffiles/041019.atsu-mar.ITSWC2004.pdf
Maruyama, A., Shibata, N., Murata, Y., Yasumoto, K. and Ito, M.: P-Tour: A Personal Navigation System for Tourism, Proceedings of 11th World Congress on ITS Nagoya, pp.18-21 (October 2004)
We propose a personal navigation system for tourism called P-Tour. When a tourist specifies multiple destinations with relative importance and restrictions on arrival/staying time, P-Tour computes the nearly best schedule to visit part of those destinations. In addition to the map-based navigation, P-Tour provides temporal guidance according to the schedule, and automatically modifies the schedule when detecting the situation that the tourist cannot follow the schedule. We have developed a route search engine as a Java Servlet which can compute a semi-optimal schedule in reasonable time using techniques of genetic algorithms.
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
(Slides) P-Tour: A Personal Navigation System for Tourist
1. P-Tour: A Personal Navigation System for Tourist Atsushi Maruyama Xanavi Informatics , Naoki Shibata Shiga University , Yoshihiro Murata Nara Institute of Sci. and Tech. , Keiichi Yasumoto Nara Institute of Sci. and Tech. , Minoru Ito Nara Institute of Sci. and Tech.
2.
3.
4.
5.
6.
7.
8.
9.
10. Route guidance mode Schedule display The entire route Moving along the scheduled route When visiting a destination. Remaining stay time/Departure time Schedule display Arrival/Departure and stay time
11.
12.
13.
14.
15.
16.
17.
18. Overview of Genetic Algorithm Candidate solutions are generated randomly National Museum 法隆寺 Horyuji Todaiji Kofukuji Yakusiji Beginning location Ending location Candidate solution GA always retains multiple candidate solutions
19. Randomly selects two solutions, and make a new solution from them Candidate solutions for the next iteration Calculate fitness values and select solutions with relatively high fitness values Repeat the iteration until predefined iteration count expires
30. Validity of output route D3 D1 D2 D7 D12 D9 D8 D10 D1 ~ D13 Importance 5 D14 ~ D30 Importance 1 Timezone D3 ≦15:00 D7 ≦19:30 Destinations in output D1, D2, D3, D7, D8, D9 D10, D12 Arrival time D3 14:50 D7 19:10 INPUT D4 D13 We changed importance to 10, and recalculated the route
31. Validity of output route D3 D1 D2 D7 D12 D9 D8 D10 D1 D2 D3 D4 D13 D5 D7 Before After D4,D13 Importance 10
Notas do Editor
Thank you. I’m Naoki Shibata. I’d like to have a talk about a navigation system suitable for tour navigation.
Our presentation consists of these topics. I first explain about the background, and then overview of P-Tour, which is the proposed system. After that, I explain the route search engine of P-Tour. And, I report results of evaluation experiments, and conclude the presentation.
Today, we have high performance PDA, small built-in GPS unit on mobile phone, Wireless LAN hotspoes, 3G mobile phone, PHS data communication card, and so on. Thanks to rapid development of these technologies, it is now realistic to implement navigation system on mobile phone or PDA. In reality, navigation service on mobile phone is already began. EZ Navi Walk is a navigation service available on mobile phone by au kddi. It uses GPS unit built-in on mobile phone, and searches for a route between two locations. It has also functions to guide users by voice and text.
But, existing navigation systems such as car navigation systems or personal navigation service by au kddi have only limited functions, which is route guidance between two locations. We believe that this is not adequate for tour navigation.
When personal navigation system is used in a tour, these situations are common. First of all, we have many destinations to visit. Also, each destination may have business hours, appointed time, and so on. We want to visit many destinations, but if we have too many destinations, we want navigation system to choose part of them. We think it would be handy if personal navigation system has these features. We propose a navigation system in which importance value and timezone can be specified for each destination. Consequently, guidance function have time schedule management.
We propose a personal navigation system for tourism, named P-Tour. The first of several features of P-Tour is tour scheduling. User inputs all destinations and corresponding timezones, importance of each destination, and Beginning and ending locations of the tour. Pre-calculation of the tour is performed according to the input data before tour starts.
And then, the system outputs route with arrival and departure time for each destination. This pre-calculation is performed within about 10 seconds.
Tour scheduling is an incremental process including several data exchanges between user and the server. User sends requests including new destinations and change of importance values for some destinations. The server performs one step of incremental calculation and returns a response within a few seconds.
Now, I explain system overview of P-Tour. Client includes route guidance program and user interface, implemented with Java MIDlet. Server includes route search engine implemented as a Java servlet. This search program utilizes map database and destination database.
These are displays at the route guidance mode. This displays the entire route. This displays arrival and departure times of each destination. This is the display on route guidance. This displays remaining stay time and departure time.
Automatic recalculation of schedule is performed according to situations. When user goes into wrong route, user’s moving speed is too slow because of traffic congestions or other reasons, or User stays at a destination too long, the system detects the situations automatically and Warns user, displays a route to return to the orignal route, or changes the schedule and the route.
The search engine have to be fast enough to realize the incremental scheduling. More concretely, The initial calculation should be performed within 10 seconds, and recalculation should be performed within a few seconds. The output route should maximize user’s satisfaction. We designed fitness function to evaluate user’s satisfaction.
User’s satisfaction should increase if many important destinations are included in the route. So, importance values of included destinations are added to the fitness value. …
Accordingly, numerical expression of the fitness function is a linear expression of the importance value for destinations included in the route, and total distance of movement. Alpha and gamma are constants. Now, I explain effect of gamma value.
Now I explain the influence of the gamma value. These are output routes with different gamma values. We can see that low gamma value leads to detour, and high gamma value leads to destinations near to the beginning location of the tour to be only selected. So, it seems to be desirable to set gamma value according to user’s preference.
Now, I explain the route search algorithm for P-Tour. P-Tour calculates route between all combination of two destinations first. A* algorithm is used for this purpose. Moving… Routes between…
Determining visiting …
GA では、まず、個体群を生成し、各個体は遺伝情報である染色体を保持し、これは解を表します。そして、交差、突然変異という演算子を用いて組み換えを行い、個体を進化させます。交差とは、染色体を切断し、組み合わせることによって新たな個体を生成します。そして、個体の評価値を計算し、選択を通じて、評価値がよくない個体を殺し、よい評価値の固体を残し、これを繰り返し、優れた解を導き出そうとするのが、遺伝的アルゴリズムです。 まず、個体群を生成し、それぞれの個体は遺伝情報をもった染色体を保持しています。そして、2つの染色体を一地点で切断し、組み合わせることで、新たな個体を生成します。新たな個体の染色体は、両方の親からの形質を受け継ぎます。そして、個体の評価し、優れた個体が生き残り、新たな親となります。これを繰り返し、染色体を進化させ、優れた評価値を探しだす手法です。
We conducted experiments to evaluate our proposed system. … We evaluated …
Calculation time and fitness values are shown here. We can realize that the fitness values converges after 50th iterations, And 10 seconds of calculation time is sufficient for practical use.
We made a program which calculates the optimal route using a branch and bound method, and compared fitness values. Calculating optimal route took 22 hours when the number of destinations is 14, and the difference of fitness values are about 1%. We believe that the approximation algorithm used in P-Tour is sufficient for practical use.
We made a program which calculates the optimal route using a branch and bound method, and compared fitness values. Calculating optimal route took 22 hours when the number of destinations is 14, and the difference of fitness values are about 1%. We believe that the approximation algorithm used in P-Tour is sufficient for practical use.
Thank you for your attention.
User’s inputs are …
We’ve conducted an experiment to evaluate recalculation of tour. The initial inputs are shown here, and the initial output route is this. This route doesn’t include destination 4 and 13, and thus we changed importance of these destinations to 10.
After recalculation, D4 and D13 are successfully included in the output route, as shown here.