path planning for a universal indoor navigation system · path planning for a universal indoor...

28
Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM Laboratory (EA 4004 CHArt) University of Paris8 France ICCHP’16, 13 – 15, July 2016 Linz, Austria

Upload: others

Post on 28-May-2020

10 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Path Planning for A Universal IndoorNavigation System

E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J.LOPEZ-KRAHE

THIM Laboratory (EA 4004 CHArt)University of Paris8

France

ICCHP’16, 13 – 15, July 2016Linz, Austria

Page 2: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Outline

1 Introduction

2 ModelingSurface ModelingUser Profile

3 Path CalculationArduousness CriterionOptimal Path Generation

4 Simulation Results

5 Conclusion

E. KAHALE et al. - ICCHP’16 2/21

Page 3: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Introduction

How can I reach mydestination ???

Airports

Commercial Centers

Railway stationsE. KAHALE et al. - ICCHP’16 3/21

Page 4: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Introduction

E. KAHALE et al. - ICCHP’16 4/21

Page 5: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Problem Statement

ObjectiveIncrease the mobility of persons in public-access building, taking into account allpotential difficulties that users might have in accomplishing different tasks ininteraction with their environment (which can be unfamiliar)

ChallengesIndoor environment

Highly distributive surfacesMulti-levels

Path PlanningInclude all personal difficulties (related to the displacement) in pathcalculationAvoiding obstacles

Proposed Solution and OriginalityUse a topological representation of the space through graph-based approachesIntroduce a new optimization criterion : the arduousnessEmploy a Universal design Concept

E. KAHALE et al. - ICCHP’16 5/21

Page 6: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Outline

1 Introduction

2 ModelingSurface ModelingUser Profile

3 Path Calculation

4 Simulation Results

5 Conclusion

E. KAHALE et al. - ICCHP’16 5/21

Page 7: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Surface Modeling

Building StructureMulti-levelsJunction elements

Each level :Many potential destinationsAmenitiesHighly distributive surfaces

Proposed ModelingTopological Representation :

Valued digraphEach walkable space is consideredas a nodeJunction elements and somespecified amenities are consideredas edges

Each node in the previousdigraph is itself a sub-digraph

Magnetic Building - Beagrenellecommercial center, Paris - France

E. KAHALE et al. - ICCHP’16 6/21

Page 8: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Surface Modeling

Building StructureMulti-levelsJunction elements

Each level :Many potential destinationsAmenitiesHighly distributive surfaces

Proposed ModelingTopological Representation :

Valued digraphEach walkable space is consideredas a nodeJunction elements and somespecified amenities are consideredas edges

Each node in the previousdigraph is itself a sub-digraph

Magnetic Building - Beagrenellecommercial center, Paris - France

E. KAHALE et al. - ICCHP’16 6/21

Page 9: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Surface ModelingSub-digraph generation

Needs

Taking into account the presence of obstacles within walkable surfaces

Proposed Solution

Apply approaches conventionally used in mobile robotics fieldsX Visibility-based method :

⇒ Two nodes share an edge if they are within line of sight ofeach other, i.e.

eij 6= ∅⇐⇒ svi + (1− s)vj ∈ Qfree ∀s ∈ [0, 1]

where Qfree denotes the walkable surfaces⇒ All points in the free space are within line of sight of at least

one node on the visibility map

E. KAHALE et al. - ICCHP’16 7/21

Page 10: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Surface ModelingSub-digraph generation

Visibility Graph Construction∗

Algorithm 1 Rotational Plane Sweep AlgorithmInput: A set of vertices Vi and a vertex vOutput: A subset of vertices from Vi that are within line

of sight of v1: For each vertex vi , calculate αi , the angle from the

horizontal axis to the line segment vvi2: Create the vertex list ε, containing the αi ’s stored in

increasing order3: Create the active list S, containing the sorted list of

edges that intersect the horizontal half-line emanatingfrom v

4: for all αi do5: if Vi is visible to v then6: Add the edge (v, vi ) to thr visibility graph7: end if8: if vi is the beginning of an edge, E , not in S then9: Insert the E into S

10: end if11: if vi is the end of an edge in S then12: Delete the edge from S13: end if14: end for

Polygonal configuration space with a start and goaland visibility connections with respect to vstart

Visibility map∗H. Choset et al., Principles of Robot Motion-Theory, Algorithms, and Implementation, The MIT Press,

Cambridge, England, 2005

E. KAHALE et al. - ICCHP’16 8/21

Page 11: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

User Profile

In French LawPeople with disabilities are classified in the following categories:

1 Physical2 Sensory3 Mental

4 Psychic5 Cognitive6 Multiple Impairment

In this workDisability = a set of difficulties to complete a task in interaction with agiven environment

E. KAHALE et al. - ICCHP’16 9/21

Page 12: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

User Profile

Pedestrian movement asks the user to:

Interact with his physical and social environmentMemorize placesEstablish connections between networksUse facilities

12 potential classes characterizing users and having an impact on thecomputation of an optimal path have been determined

Related to the use of hands◦ Trembling or involuntary movements◦ Difficult use of hands

Related to the displacement

◦ Supporting stick, crutches, walker◦ Stiffness, joint or muscle pains◦ Shortness of breath, respiratory or heart problems◦ Electric wheelchair◦ Manual wheelchair

Related to the vision

◦ Blindness (use of a walking cane)◦ Glare, fog, blur, opacity, paleness◦ Fragmented, Tunnel or peripheral vision◦ Blindness (use of guide dog)◦ Difficulty reading

E. KAHALE et al. - ICCHP’16 10/21

Page 13: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Outline

1 Introduction

2 Modeling

3 Path CalculationArduousness CriterionOptimal Path Generation

4 Simulation Results

5 Conclusion

E. KAHALE et al. - ICCHP’16 10/21

Page 14: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Arduousness Criterion

NeedsProvided path must:

Satisfy the physical capacities of the userMinimize the effort needed to accomplish each step

PropositionIntroduce a new coefficient describing the arduousness associatedto each edge in the digraph

E. KAHALE et al. - ICCHP’16 11/21

Page 15: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Arduousness Criterion

NeedsProvided path must:

Satisfy the physical capacities of the userMinimize the effort needed to accomplish each step

PropositionIntroduce a new coefficient describing the arduousness associatedto each edge in the digraph

E. KAHALE et al. - ICCHP’16 11/21

Page 16: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Arduousness Criterion

Determine the inherent characteristics for each amenityidentified as an edgeExample:

Inherent characteristic for a stairStaircase openwork (without riser)

With two flightsWith three flights

SpiralWithout right handrail, unusable left handWithout left handrail, unusable right hand

Without handrailShortLong

Down stairUp stair

Without orientation (Bi-directional)Presence of palisade narrowing the width of the stairPresence of scaffold narrowing the width of the stair

Presence of barrier blocking access

E. KAHALE et al. - ICCHP’16 12/21

Page 17: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Arduousness Criterion

For each property we define a weighting coefficient γu,ei So theglobal weight for the edge ei is given by:

Γei =∏

γu,ei

where γu,ei ∈ [0, 1] ⇒ Γei ∈ [0, 1], and

Γei =

{0; Impassable1; No constraint

E. KAHALE et al. - ICCHP’16 13/21

Page 18: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Arduousness CriterionExample

StairInherent characteristic Weight User’s Difficulties Weight

Staircase openwork (without riser) 0,9 Rolling bulky object 0,8With two flights 0,9 Carrying bulky object 0,8With three flights 0,8 Trembling or involuntary movements (upper

limbs)0,8

Spiral 0,8 Difficult use of hands 0,7Without right handrail, unusable left hand 0,9 Blindness (use of a walking cane) 0,9Without left handrail, unusable right hand 0,9 Glare, fog, blur, opacity, paleness 0,9Without handrail 0,9 Stiffness, joint or muscle pains (lower limbs) 0,7Short 1 Fragmented, Tunnel or peripheral vision 0,9Long 0,9 Difficulty reading 1Down stair 1 Blindness (use of guide dog) 1Up stair 0,9 Supporting stick, crutches, walker 0,6Without orientation (Bi-directional) 0,9 Shortness of breath, respiratory or heart problems 0,8Presence of palisade narrowing the width of the stair 0,95 Electric wheelchair 0Presence of scaffold narrowing the width of the stair 0,95 Manual wheelchair 0Presence of barrier blocking access 0

The arduousness coefficient of a stair WITHOUT HANDRAIL, SHORT et GOING DOWN for a person havingHEART PROBLEMS and CARRYING BULKY OBJECT is given by:

Γei = 0.9× 1× 1× 0.8× 0.8 = 0.9× 0.64 = 0.576

E. KAHALE et al. - ICCHP’16 14/21

Page 19: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Optimal Path Generation

NeedsProvided path must:X Satisfy the physical capacities of the user

Minimize the effort needed to accomplish each step

Problem StatementCombinatorial Optimization Problem / Operational ResearchShortest Path Problem

Proposed SolutionUse a path planning method conventionally applied in mobilerobots navigation systems

E. KAHALE et al. - ICCHP’16 15/21

Page 20: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Optimal Path Generation

Dijkstra’s Algorithm† ‡ § ¶

X Find the shortest pathX Ensure the optimality of the solution

X Replace the distance by 1/Γei ⇒ Maximize Γei ⇔ Providespath with the minimum arduousness cost

†H. Choset et al. Principles of Robot Motion-Theory, Algorithms, andImplementation, The MIT Press, Cambridge, England, 2005‡S.M. LaValle, Planning Algorithms, Cambridge University Press, U.K.:

Cambridge, 2006§B. Siciliano et al. Robotics: Modelling, Planning and Control, Springer Verlag,

London, 2009¶D. Jungnickel, Graphs, Networks and Algorithms. Fourth Edition, Springer

Verlag, London, 2013E. KAHALE et al. - ICCHP’16 16/21

Page 21: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Optimal Path Generation

Dijkstra’s Algorithm† ‡ § ¶

X Find the shortest pathX Ensure the optimality of the solution

X Replace the distance by 1/Γei ⇒ Maximize Γei ⇔ Providespath with the minimum arduousness cost

†H. Choset et al. Principles of Robot Motion-Theory, Algorithms, andImplementation, The MIT Press, Cambridge, England, 2005‡S.M. LaValle, Planning Algorithms, Cambridge University Press, U.K.:

Cambridge, 2006§B. Siciliano et al. Robotics: Modelling, Planning and Control, Springer Verlag,

London, 2009¶D. Jungnickel, Graphs, Networks and Algorithms. Fourth Edition, Springer

Verlag, London, 2013E. KAHALE et al. - ICCHP’16 16/21

Page 22: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Outline

1 Introduction

2 Modeling

3 Path Calculation

4 Simulation Results

5 Conclusion

E. KAHALE et al. - ICCHP’16 16/21

Page 23: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Simulation Results

Proof of concept application has been :developed in JAVAintegrated in a Samsung Galaxy S6 (64-bit processor)

Increase the robustness facing map data collection errorsData format or unit errorsImprecision in the location of points of interest (outsidewalkable surfaces)

Two scenarios were chosen

E. KAHALE et al. - ICCHP’16 17/21

Page 24: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Simulation ResultsScenario 1

Scenario 1 : Person without difficulties

(a) Departing point → Escalator (b) Escalator → Arrival point

E. KAHALE et al. - ICCHP’16 18/21

Page 25: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Simulation ResultsScenario 2

Scenario 2 : Wheelchair user

(a) Departing point → Lift (b) Lift → Arrival point

E. KAHALE et al. - ICCHP’16 19/21

Page 26: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Outline

1 Introduction

2 Modeling

3 Path Calculation

4 Simulation Results

5 Conclusion

E. KAHALE et al. - ICCHP’16 19/21

Page 27: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Conclusion and Future Works

Novel path planning strategy for indoor navigation system based on a universaldesign concept

Surface ModelingTopological representation (digraph)Highly distributive surfaces (family of paths, obstacle avoidance)Determine the inherent characteristics of each amenity identified as anedge

User Profile

Identify 12 potential difficulties having an impact on the displacement

Optimal path generationIntroduce new criterion : Arduousness for optimizationMinimizing Arduousness : Dijkstra’s Algorithm

Validating through Simulation

Experimentation in a large Parisian railway (in discussion)

E. KAHALE et al. - ICCHP’16 20/21

Page 28: Path Planning for A Universal Indoor Navigation System · Path Planning for A Universal Indoor Navigation System E. Kahale, P. C. Hanse, V. DESTIN, G. UZAN, and J. LOPEZ-KRAHE THIM

Introduction Modeling Path Calculation Simulation Results Conclusion

Thank You!

E. KAHALE et al. - ICCHP’16 21/21