We present a novel user interface concept for indoor navigation which uses directional arrows and panorama images at decision points. The interface supports the mental model of landmark-based navigation, can be used on- and offline and is highly tolerant to localization inaccuracy.
2. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Overview
Motivation: Vision-Based Localization
UI Concept and First Evaluation
Decision-Point Based Navigation
User Study
Discussion and Future Work
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
3. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Background and Motivation
§ Location information still the
most important contextual information
§ Indoor localization is a hot topic
and useful for a lot of locations:
□ Airports
□ Hospitals
□ Conference venues
□ Large environments
§ Various indoor localization methods
possible:
□ WLAN/cell-based localization
□ Sensor-based localization
□ Beacon-based localization
□ Vision-based localization
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
4. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Vision-Based Localization Live features
§ Principle
□ Comparing of query images, taken with
camera, to reference images based on
visual features
§ Advantages
□ No infrastructure and augmentation of the environment needed
□ Immune to disturbances (reflection/refraction…)
□ Works with existing
(modern) hardware
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
5. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Overview
Motivation: Vision-Based Localization
UI Concept and First Evaluation
Decision-Point Based Navigation
User Study
Discussion and Future Work
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
6. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
HCI Perspective: Interaction Concept
§ Augmented Reality View for intuitiveness
□ But: Consequence when location estimate
is inaccurate:
wrong overlays!
?
§ Permanent re-localization during path required
□ High computational effort
□ Uncomfortable camera pose
□ Not all scenes adequate for visual localization
(sometimes insufficient features for algorithm)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
7. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Interaction Concept (cont.)
Instead:
§ Re-localization from time to time
(when possible)
□ Estimation of intermediate position
with odometry (device sensors)
□ More comfortable usage
§ “Virtual Reality” View
□ Regularly show 360° panorama images
of the environment with embedded navigation
instructions (those panoramas are already available
from reference images)
□ More robust: user matches virtual and real world
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
8. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Initial User Study
§ Comparison of
Augmented Reality and
Virtual Reality
§ Simulation of different levels
of accuracy
§ Survey of user preferences
See Full Paper:
A. Möller, M. Kranz, R. Huitl, S. Diewald, L. Roalter
A Mobile Indoor Navigation System Interface
Adapted to Vision-Based Localization
In: 11th Intl. Conf. on Mobile and
Ubiquitous Multimedia (MUM 2012)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
9. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Initial User Study: Summary
§ Panoramas provide better guidance when location estimate is
inaccurate than Augmented Reality
§ But: Subjects experienced frequent changes (~2m) as irritating, as
□ Shown locations are not always correct
□ Locations sometimes do not look like in reality
□ Frequent changes in the UI are visually straining
§ Subjects manage finding their goal, also when not looking at the
screen all the time
§ How to improve?
□ Reduce frequent visual changes
□ Simplify match between panoramas and real world (problem of
misplaced panoramas?)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
10. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Overview
Motivation: Vision-Based Localization
UI Concept and First Evaluation
Decision-Point Based Navigation
User Study
Discussion and Future Work
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
11. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible (by swiping through list of route instructions)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
12. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
13. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
14. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
15. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
16. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
17. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Decision-Point Based Navigation (DPBN)
§ Show only panoramas of decision points
§ Compliance with familiar mental model of self-orientation
§ Users can always re-localize and retrieve the panorama
of their current location
§ High error tolerance
§ Offline usage possible
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
18. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Overview
Motivation: Vision-Based Localization
UI Concept and First Evaluation
Decision-Point Based Navigation
User Study
Discussion and Future Work
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
19. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
User Study: Research Questions
§ RQ1: Does DPBN have an effect on efficiency?
Are users as fast as with continuous panoramas?
§ RQ2: Is DPBN more convenient than continuous panoramas?
What mode do users prefer? How well do they feel guided?
§ RQ3: What usage patterns can be identified?
What do we learn for designing an “ideal” route description?
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
20. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
User Study – Quick Facts
3 conditions
- Continuous panoramas (automatic *) * Wizard of Oz
- Decision points (automatic *)
- Decision points (manual navigation)
3 routes (avoidance of learning effects)
Measured data: time, user behavior, questionnaires
12 Participants, 75% male, 25% female, average 25 years
Within-subjects design
75% own a smartphone, only one has experience with
indoor navigation
Source: iconsdb.com
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
21. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ1
Average Time per Mode
350
300
Time in seconds
250
200
Difference
150 not
significant 263
100 196 208
50
0
Continuous Decision-Point Decision-Point
(automatic) (automatic) (manual)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
22. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ2
I found the method pleasing to use.
5: fully agree
1: fully disagree
- Median
+ Mean
Continuous Decision-Point Decision-Point
(automatic) (automatic) (manual)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
23. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ2
I felt guided well to the goal.
5: fully agree
1: fully disagree
- Median
+ Mean
Continuous Decision-Point Decision-Point
(automatic) (automatic) (manual)
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
24. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ2
Panoramas of decision
points are sufficient for Which system would you choose?
orientating
8%
Continuous
17% (automatic)
Decision-Point
(automatic)
Decision-Point
75% (manual)
5: fully agree
1: fully disagree
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
25. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ3
Participant 12 - Path 1
40
30
Distance (m)
20
10
0
20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340
Time (s)
swipe relocalize Δd position
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
26. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ3
Participant 2 - Path 2
60
50
40
30
20
10
Distance (m)
0
20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340
-10
-20
-30
-40
-50
-60
Time (s)
swipe relocalize Δd position
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
27. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ3
Participant 5 - Path 3
30
20
Distance (m)
10
0
20 40 60 80 100 120 140 160 180 200 220
-10
-20
Time (s)
swipe relocalize Δd position
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
28. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Results: RQ3 – Manual Mode
§ Average numbers
□ 52.5 continuous locations (automatic, continuous mode)
□ 10.3 decision points (automatic, decision-point mode)
□ 15.3 manual swipes
□ 9.3 times relocalized
§ Different usage patterns/strategies
□ Showing “ahead” panoramas, no use of automatic re-localization
□ Frequent use of re-localization, (sometimes only in parts of the
route) – equivalent to continuous panorama mode
□ “Memorizing” of the entire route not observed, subjects only
looked at the “next” panorama location
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
29. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Overview
Motivation: Vision-Based Localization
UI Concept and First Evaluation
Decision-Point Based Navigation
User Study
Discussion and Future Work
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
30. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Discussion
§ Subjects were equally fast with DPBN as
with continuous panoramas (RQ1)
§ Subjects felt guided better with continuous panoramas (RQ2)
□ Due to more certainty whether they are on the right way
□ Confirmed by frequent swipes and re-localizations in manual mode
§ DPBN is desirable from a technical point of view (reliability…), but
acceptance must be increased
§ Requires stronger confirmation whether subjects are (still) on the right
way
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
31. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
How could this be achieved?
(Lessons learned)
§ Compromise between continuous mode and DPBN as it is now
§ Distance information
§ Verification when decision point was passed
§ Add intermediate landmarks in decision point mode
□ Especially if decision point lies too far away (distance threshold)
§ Simplify the mapping of picture content to real environment
□ Highlight significant objects in panorama
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
32. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Summary and Outlook
§ With Decision-Point Based Navigation, we have presented
a novel interface for visual indoor navigation
□ Very robust in case of localization inaccuracy
□ Works even if localization fails (manual mode)
□ Supports users’ mental model of navigation (landmarks)
§ Real-world study results
□ DPBN as efficient as full panorama navigation
□ DPBN needs more information to reduce users’ uncertainty
(landmarks, confirmations) to increase satisfaction
§ Future work
□ See previous slide ;-)
□ Identification of good candidates for (additional) key points
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
33. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Thank you for your attention!
Questions?
?
?
andreas.moeller@tum.de
www.vmi.ei.tum.de/team/andreas-moeller.html
This research project has been supported by the space agency of the German Aerospace Center with
funds from the Federal Ministry of Economics and Technology on the basis of a resolution of the German
Bundestag under the reference 50NA1107.
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
34. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
Paper Reference
§ Please find the associated paper at:
https://vmi.lmt.ei.tum.de/publications/2013/MCPT2013-
IndoorNav_preprint.pdf
§ Please cite this work as follows:
§ A. Möller, M. Kranz, L. Roalter, S. Diewald, Kåre Synnes
Decision-Point Panorama-Based Indoor Navigation
In: 14th International Conference on Computer Aided Systems
Theory (EUROCAST 2013), pp. 325-326, Las Palmas de Gran
Canaria, Spain, February 2013
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes
35. Institute for Media Technology
Distributed Multimodal Information Processing Group Technische Universität München
If you use BibTex, please use the following
entry to cite this work:
@INPROCEEDINGS{MCPT13IndoorNav,
author = {Andreas M{"o}ller and Matthias Kranz and Luis Roalter and Stefan Diewald},
title = {{Decision-Point Panorama-Based Indoor Navigation}},
booktitle = {14th International Conference on Computer Aided Systems Theory (EUROCAST 2013)},
editor = {Alexis Quesada-Arencibia and Jos'{e} Carlos Rodriguez and Roberto Moreno-Diaz jr. and Roberto Moreno-Diaz},
year = {2013},
month = feb,
pages = {325--326},
ISBN = {978-84-695-6971-9},
location = {Las Palmas de Gran Canaria, Spain},
}
15.2.2013 A. Möller, M. Kranz, L. Roalter, S. Diewald, K. Synnes