Proactive Navigation Optimized for Individual Users

J. Ikedo, Tsutomu Horioka, Yasuhiro Niikura, Yoshimasa Koike, Hiroshi Sawada, Yuichi MutoNTT 2015 · DOI

Abstract

In this article, we first give an overview of the proactive navigation project that NTT Service Evolution Laboratories is striving to accomplish. Then we provide a simple explanation of the technologies currently being researched and developed to implement the project: information navigation, big data analysis, and resilient communication. Finally, we give an overview of experiments currently being run in Fukuoka City along with their results.

Introduction

Standard navigation systems have traditionally operated as reactive tools, responding to specific user inputs to provide routes from point A to point B. However, this model often lacks contextual intelligence and fails in environments where GPS is unreliable or during natural disasters when cloud connectivity is severed. Researchers at NTT Service Evolution Laboratories are redefining this paradigm through "Proactive Navigation."

Inspired by the Japanese spirit of omotenashi (anticipatory hospitality), this framework aims to provide relevant information before a user explicitly requests it. By integrating advanced machine learning with resilient networking, the system seeks to enhance both the daily convenience and the emergency safety of urban travelers.

Technical Framework

The proactive navigation project is built upon three core technological pillars: Information Navigation, Big Data Analysis, and Resilient Communication.

Information Navigation: Precision and Context

To provide proactive suggestions, the system must first understand a user's current state with high precision. Standard GPS often suffers from significant margins of error in dense urban "canyons" or underground spaces. To mitigate this, the researchers utilize spatiotemporal clustering. This technique identifies "retention points"—locations where a user remains for a significant duration—and matches these patterns against known points of interest (POIs).

Furthermore, the system employs Deep Neural Networks (DNNs) for transport mode estimation. Unlike simple speed-based algorithms, these models can distinguish between walking, buses, trains, and taxis by recognizing complex movement signatures, even in heavy traffic.

Technical workflow for location and transport mode estimation using spatiotemporal clustering and deep learning.

Big Data Analysis and Prediction

The second pillar involves analyzing multidimensional composite data to predict future states. This includes forecasting crowd density to prevent "crushes" and clustering users into specific personas based on age, gender, and nationality. By understanding these clusters, the system can tailor recommendations—such as suggesting a scenic walk to a tourist who has just finished a meal—thereby shifting the user interface from search-based to discovery-based.

Resilient Communication

Safety is a critical component of hospitality. In the event of a disaster where the internet is disconnected, the system transitions to a browser-based mesh network. Utilizing HTML5 and local Wi-Fi access points, user devices can communicate "hop-by-hop." In this model, individuals moving between connectivity zones act as physical data relays, bridging isolated areas to ensure that emergency information can still reach evacuees without a global web connection.

Experimental Results: The Fukuoka City Trial

The effectiveness of these technologies was evaluated during a regional revitalization trial in Fukuoka City, launched in October 2014. The trial collected GPS logs and usage history from foreign and domestic tourists to identify behavioral patterns across the city.

The results demonstrated that spatiotemporal clustering significantly improved positioning accuracy in dense urban environments. Moreover, the system successfully identified distinct tourist clusters, such as "40s Asian males" interested in nature and hot springs in the Tenjin-Minami area. These insights allow for highly targeted proactive recommendations that align with specific user interests.

Tourist clusters and personas identified in the Fukuoka City trial, highlighting specific demographic interests in Nakasu, Tenjin-Minami, and Dazaifu.

Discussion and Future Outlook

While the Fukuoka trial proved the feasibility of proactive navigation, several challenges remain. One significant conceptual hurdle is the "Intervention Loop" paradox: if a system successfully predicts and prevents a crowd crush by redirecting users, the original prediction becomes technically "inaccurate." Managing this real-time influence on human behavior requires continuous, low-latency computational iteration.

Future work will focus on closing the "prediction gap" to provide high-confidence forecasts for individual users rather than just demographic clusters. Achieving true "digital hospitality" will ultimately require a collaborative ecosystem, integrating diverse data sources and service providers to create a seamless, anticipatory travel experience.

Summary updated

Abstract

In this article, we first give an overview of the proactive navigation project that NTT Service Evolution Laboratories is striving to accomplish.Then we provide a simple explanation of the technologies currently being researched and developed to implement the project: information navigation, big data analysis, and resilient communication.Finally, we give an overview of experiments currently being run in Fukuoka City along with their results.
BibTeX
@inproceedings{ikedo2015proactive,
  author = {J. Ikedo and Tsutomu Horioka and Yasuhiro Niikura and Yoshimasa Koike and Hiroshi Sawada and Yuichi Muto},
  title = {{Proactive Navigation Optimized for Individual Users}},
  booktitle = {NTT},
  year = {2015},
  doi = {10.53829/ntr201507fa4},
}
Figure