An experimental study in using natural admixture as an alternative for chemic...
Master Thesis 2015 - Aykut İçöz at University Of Ulm
1. Fie1
Presentation of Master Thesis
Development of a Dialog-based Pedestrian
Navigation System
Master Thesis Topic
M.Sc. Student : Aykut Icoez
Supervisor : Dr. Michael Angermann
2. 2
Content
• Objective
• Information about Pedestrians
• Equipment used and Experimental Setup
• Photos of Equipment and Running Application
• Overview of a Dialog System
• Main Parts of a Dialog -with an example-
• Landmarks
• Availability and Visibility : criteria
• Dijkstra’s Algorithm: shortest path
• Sequence of Events in a general application
• Tests and Results
• Outlook
3. 3
Objective
Navigation
System
Dialog
SystemInterface
Main part to develop in this work Preparation for further work
Subject application : Navigation system for pedestrians
Two black boxes : The dialog system and the application
Interface : Interaction of both according to requirements
4. 4
Background Information about Pedestrians
• Characteristics of Pedestrian Movement:
– Movement Freedom
• Principally all ways free (streets + footpaths)
• Not bounded to certain paths
– Speed
• 5 km/h (1.5 m/s)
– Field of vision
• Orientation in movement direction
• Principally +/- 180 Deg. in line of sight
– Enough time to observe/sense surrounding environment
• Guidance of Pedestrians:
– Requires personalization due to the huge variety of different user demands
– Must be subdivided into groups according to user tasks and information needs
• Pedestrian types:
– tourist, business traveler, resident, disabled person
• Most appropriate type of route - Routing:
– shortest distance, least time, fewest turns, one with most point of interests (POI)
9. 9
View of the Pedestrian Representation
Inputs: Position (GPS) and Direction of View (Compass)
Green dot: Pedestrian (user)
10. 10
Overview of a Dialog System
• Speech Recognizer : produces a word lattice (sequence of words)
• Semantic Parser : determines the meaning of the query, produces best interpretation
(extracting a meaning)
• Dialog Manager : resolve ambiguities, clarifies with user, builds database queries
(initiating system actions)
• Database interface : uses the meaning representation and to generate a database
query & to access database
• Response generator : presents the interaction result in the form of text, speech,
tables or graphics.
11. 11
Main Parts of a Dialog
• First Part:
– Aim : Target spefication (e.g.: go, reach to…)
– Direction : From the user to the application
• Second Part:
– Aim : Orientation (e.g.: stand with your back to, turn around…)
– Direction : From the application to the user
• Third Part:
– Aim : Moving (e.g.: walk, go, continue straight/towards, across…)
– Direction : From the application to the user
• Fourth Part:
– Aim : Arrival (e.g.:arrive at, reach to, find…)
– Direction : From the application to the user
12. 12
1. Part : Target spefication
Computer: Where do you want to go?
User : I want to go to <library>. (<casino> or <canteen>)
Computer: To <library>?
User : <Yes>/<No>
2. Part : Orientation
Computer: Stand with your back to Institute <A>. (<B>, <C>, <D>)
Computer: Turn <left> around yourself. (<right>)
3. Part : Moving
Computer: Walk (<…> meters) straight till you reach Institute <A> at your <left>.
Computer: Do you see Institute <A> at your <left>?
User : <Yes>/<No>
Computer: Turn <left>. Walk (<…> meters) straight till you reach <casino> at your <right>.
4. Part : Arrival
Computer: You have arrived at <library>.
Example of a Dialog
Landmarks : Institute <A> or <B> or <C> or <D>
Alternatives : are shown with <…>.
Method used : System-directed, system asks all the questions, user is passive.
13. 13
Landmarks
• Definition:
– Every orientation point in the range of movement
– Used as a reference
• Features:
– Special visual characteristics (noticeable difference to neighborhood)
– A special function/meaning or central state/location
• Reasonable locations in practice:
– Along the path / at decision points / far away, but visible
• Possible types:
– Architecture (towers, main entrance, monuments, columns, bridges)
– Commercial activity (newspaper kiosk, fish market)
– Buildings (institutions, schools, fire stations, churches)
– Street-network (crossroads, junctions)
– Water features (rivers, channels)
– Various (telephone booths, parking lots)
Note: Colorful ones show the ones chosen at the DLR site.
17. 17
Graphical View of Landmarks at DLR Site
Total number of chosen landmarks: 20
Two way
One way
18. 18
Photos from Landmark 1 and 2
At Landmark 1 (LM1) At Landmark 1 (LM1)
At Landmark 2 (LM2) At Landmark 2 (LM2)
LM2
LM20 LM19
LM1
LM3
19. 19
Availability and Visibility
• Two criteria used to identify relation between different landmarks
• Availability:
– A route connects the source LM to the next LM lying in the neighborhood.
– There is no other LM lying in-between.
• Visibility:
– A human can see the next LM as he stands at the source LM.
• Examples at DLR site:
– Available / Visible (A / V) : LM1 – LM2
• Connecting route without any other LM on the way and no obstacle in-between
– Not available / Visible(NA / V) : LM1 – LM19
• Connecting route with another LM (number 20) on the way and no obstacle in-between
– Available / Not visible(A / NV) : LM2 – LM8
• Connecting route, however obstacle blocking (Building 110)
– Not available / Not visible (NA / NV) : LM1 – LM8
• No direct route connection and obstacles blocking line of sight
21. 21
…A / V A / VA / V NA / VA / VLM 20
A / V…A / NV NA / VNA / VLM 19
A / NV…A / V LM 18
A / V A / V…A / V LM 17
A / V A / V…A / V LM 16
A / V…A / VA / V A / NV LM 15
A / V…A / NV A / NVA / NV LM 14
A / VA / NV…A / V LM 13
A / V…A / V A / NVLM 12
A / V…A / V A / VA / NVLM 11
A / NV A / V…A / V LM 10
A / NVA / NV A / V…A / NV A / NV LM 9
A / NV…A / NV A / NVA / NVA / V LM 8
A / NV…A / V LM 7
A / V…A / NVA / V LM 6
A / NV…A / V LM 5
A / NVA / NV A / VA / V…A / V LM 4
A / NV A / V…A / V LM 3
NA / VNA / V A / NV A / V…A / VLM 2
A / VNA / V A / NVA / NV A / V…LM 1
LM
20
LM
19
LM
18
LM
17
LM
16
LM
15
LM
14
LM
13
LM
12
LM
11
LM
10
LM 9LM 8LM 7LM 6LM 5LM 4LM 3LM 2LM 1
Adjacency Matrix of Landmarks at DLR
NA / NVA: available, V: visible, NA: not available, NV: not visible
23. 23
Dijkstra’s Algorithm
• Input: Routing matrix
• Purpose:
– To find the shortest paths from a single source vertex to all
other vertices in a weighted, directed graph
• Structure:
– Begins at a specific vertex and extends outward within the
graph, until all vertices have been reached.
– Stores total distance from a source vertex to the current vertex
• Status in Our System:
– Implemented and tested succesfully, however not integrated yet
26. 26
Tests and Results
• WoZ experiments were executed with an experienced
and another unexperienced user with success.
– WoZ: The user, without being aware, in interaction with
the administrator who simulates the system
• Corrections and fine-tuning were made considering
landmark positions and their suitability.
• Targeted Navigation Application to interact with a
future Dialogue System is obtained.
27. 27
Outlook
• A positioning system is developed and available that is able to
guide a user to reach his destination via the shortest path.
• As a prediction, possible works in future could include the
following:
– The tests can be carried out and extended to a real life
situation by starting at the Landsberg testbed of DLR.
– An appropriate Dialog Strategy could be developed and
used with the Application.
– A Dialog System (possibly with an interface) developed
according to requirements could be implemented with its
main parts interacting with the Navigation Application .
28. 28
Fie28
Thank you for your attendance.Thank you for your attendance.
Any questions?
Comments are very welcome.
