Yi Ding

Assistant Professor at UT Dallas

Key Words: Indoor Navigation

Taxonomy

  • Sensor: Wi-Fi, IMU, Camera, Geomagnetic, LiDAR;
  • Floorplan-free / Floorplan-based;

Comments

  • “Localization” and “Navigation” are two closely-related topics.
  • The basic idea of Travi-Navi, FOLLOWME, and ppNav are similar in terms of the “Guider+Follower” scheme, the primary difference comes in data (sensor) source.
  • The basic weaknesses of existing works (Travi-Navi, FOLLOWME, and ppNav) are the need of guider (leader) and high-sampling sensor data for trace construction, deviation detection or turn detection.

Papers

  • Travi-Navi, Self-deployable Indoor Navigation System, [MobiCom-2014-Travi-Navi]
    • Taxonomy: Camera, Floorplan-free;
    • Idea: (1) Enable user to bootstrap simple navigation service; (2) Guider trace record + Follower scheme.
    • Contribution: (1) Idea: Guider+Follower; (2) Tech: Sensor fusion, shortcut discovery, image blur process.
    • Weaknesses: (1) Images unavailable for many applications; (2) Need guider;
  • Last-mile navigation using smartphones, [MobiCom-2015-FOLLOWME]
    • Taxonomy: Geomagnetic, Floorplan-free;
    • Idea: (1) Guider+Follower scheme; (2) Geomagnetic fingerprints;
    • Contribution: (1) Idea: Infra-free; (2) Tech: Magnetic-field-based, step-constrained trace synchronization; (3) Implementation
    • Weaknesses: (1) Need high sampling rate (~100Hz) and energy consumption.
  • ppNav, Peer-to-Peer Indoor Navigation, [IEEE-2017-P2P]
    • Taxonomy: Wi-Fi+IMU, Floorplan-free;
    • Idea: Navigation according to a previous traveler’s trace experience.
    • Contribution: (1) Idea: Infra-free, P2P navigation; (2) Tech: WiFi fingerprint model: diagrammed; (3) Implementation.
    • Weaknesses: (1) Need guider or labeled data; (2) Recording events of interests (heading, turning, upstairs/downstairs) need high sampling rate (~100Hz) and energy consumption.
  • Deep Learning-based Wireless Localization for Indoor Navigation, [MobiCom-2020-DLoc]
    • Taxonomy: Wi-Fi, LiDAR;
    • Idea: (1) Use the neural network to implicitly model environment (for Wi-Fi-based localization);
    • Contribution: (1) Domain knowledge+NN to solve Domain-specific problem; (2) Map construction via robot; (3) Large-scale data-set.
    • Weaknesses: (1) Need map contribution (fingerprinting) process (MapFinder, SLAM).

Ref.

[MobiCom-2014-Travi-Navi] Zheng, Yuanqing, Guobin Shen, Liqun Li, Chunshui Zhao, Mo Li, and Feng Zhao. “Travi-navi: Self-deployable indoor navigation system.” IEEE/ACM transactions on networking 25, no. 5 (2017): 2655-2669.

[MobiCom-2015-FOLLOWME] Shu, Yuanchao, Kang G. Shin, Tian He, and Jiming Chen. “Last-mile navigation using smartphones.” In ACM MobiCom, pp. 512-524. 2015.

[IEEE-2017-P2P] Yin, Zuwei, Chenshu Wu, Zheng Yang, and Yunhao Liu. “Peer-to-peer indoor navigation using smartphones.” IEEE Journal on Selected Areas in Communications 35, no. 5 (2017): 1141-1153.

[MobiCom-2020-DLoc] Ayyalasomayajula, Roshan, Aditya Arun, Chenfeng Wu, Sanatan Sharma, Abhishek Rajkumar Sethi, Deepak Vasisht, and Dinesh Bharadia. “Deep learning based wireless localization for indoor navigation.” In ACM MobiCom, pp. 1-14. 2020.