Timeliness-aware On-site Planning Method for Tour Navigation

TL;DR

This paper introduces a timeliness-aware on-site tour planning method that dynamically optimizes tourist routes considering real-time factors, improving decision-making during visits with fast algorithms tested in Kyoto.

Contribution

It proposes a novel tour score model incorporating static and dynamic factors and develops three greedy algorithms for efficient on-site route optimization.

Findings

Algorithms achieved solutions within seconds, suitable for real-time use.

Proposed methods outperformed baseline routes in Kyoto case study.

Dynamic factors significantly improve tour satisfaction and resource utilization.

Abstract

In recent years, there has been a growing interest in travel applications that provide on-site personalized tourist spot recommendations. While generally helpful, most available options offer choices based solely on static information on places of interest without consideration of such dynamic factors as weather, time of day, and congestion, and with a focus on helping the tourist decide what single spot to visit next. Such limitations may prevent visitors from optimizing the use of their limited resources (i.e., time and money). Some existing studies allow users to calculate a semi-optimal tour visiting multiple spots in advance, but their on-site use is difficult due to the large computation time, no consideration of dynamic factors, etc. To deal with this situation, we formulate a tour score approach with three components: static tourist information on the next spot to visit, dynamic tourist information on the next spot to visit, and an aggregate measure of satisfaction associated with visiting the next spot and the set of subsequent spots to be visited. Determining the tour route that produces the best overall tour score is an NP-hard problem for which we propose three algorithms on the greedy method. To validate the usefulness of the proposed approach, we applied the three algorithms to 20 points of interest in Higashiyama, Kyoto, Japan, and confirmed that the output solution was superior to the model route for Kyoto, with computation times of the three algorithms of $1.9\pm0.1$, $2.0\pm0.1$, and $27.0\pm1.8$ s.