一個路徑推薦導向的動物園導覽系統

在本研究中，我們為動物園的遊客提出了一個嶄新的大規模人群導航系統。該系統運用了機器學習的方式分析遊客的興趣，為每位遊客提供動態的路徑推薦服務。通過利用智能手機的便利性，系統可以簡單地收集遊客的興趣與基本資料。最後，根據動物園當前的人群分佈、遊客的問卷回饋以及遊客的GPS歷史記錄，系統為每位訪客或團體動態地計算合適的路線並推薦給他們。在我們的系統中採用的機器學習技術包含了K-Means的Clustering、協同過濾推薦與馬可夫鏈蒙地卡羅法三種演算法，來計算使用者的興趣資訊給伺服器做路徑推薦，並且定時的更新園區狀況與遊客狀態，以達到更為正確的路徑推薦結果。本系統透過一個類似Pokémon GO的遊戲，以動物肖像變形技術與遊戲化後的人群疏散系統做為機關巧思的基礎。與傳統的遊園方式相比，遊客可以在動物園玩遊戲，獲得更多有趣的回饋和體驗。遊客們可以通過智慧型手機上的遊戲 App收到伺服器發出的遊戲指示，並且移動到特定位置並回答小遊戲的提問，便可以累積積分，同時蒐集生動有趣的個人專屬動物彩妝照。整合以上技術，我們不但可以提供遊客們一個寓教於樂的系統，同時還可以減少動物園擁擠的情形。

關鍵字

機械學習；人群導向；遊戲化；仕掛學；人群模擬

並列摘要

In this study, we presented a new large-scale crowd navigation system for visitors to the zoo. The system uses machine learning to analyze the interests of visitors and provides a dynamic path recommendation service for each visitor. By taking advantage of the convenience of smartphones, the system can simply collect the tourist's interests and basic information. Finally, based on the zoo’s current population distribution, visitor feedback, and visitor’s GPS history, the system dynamically calculates and recommends suitable routes for each visitor or group. The machine learning techniques used in our system include K-Means Clustering, Collaborative Filtering Recommendations, and Markov Chain Monte Carlo Method to calculate user interest information to the server for path recommendation. And regularly update the status of the zoo and the status of visitors in order to achieve a more correct path recommended results. Through a game similar to Pokémon GO, the system uses animal portrait morphing technology and gamified crowd evacuation system as the basis of the ingenuity. Compared with traditional zoo-visiting, visitors can play games at the zoo and get more interesting feedback and experience. Visitors can receive the game quest from the server through the game app on the smartphone, and move to a specific location and answer questions from the mini-games to get game points and collect vivid, interesting personal morphing portrait. By integrating the above technologies, our system not only provides visitors with an edutainment system but also reduce the crowdedness of the zoo.

並列關鍵字

Machine Learning ； Crowd Navigation ； Gamification ； Shikakeology ； Crowd Simulation

參考文獻

一、論文

Google Scholar

[1] Xiaogang Jin, Jiayi Xu, Charlie C.L. Wang, Shengsheng Huang, Jun Zhang, “Interactive Control of Large Crowd Navigation in Virtual Environment Using Vector Field”, 2008.