Driver Model: Simulation Development and Data Integration

1
Driver Model Simulation Development and Data
Integration

• Delphine Cody Swe-Kuang Tan
• Path Conference

2
Background

• Integrate a driver model within SmartAHS
• MOU 369 from a vehicle control to a driver
  control perspective
• creation of an architecture of drivers cognition
  and perception
• Application of the architecture to highway
  driving, mainly car following cases
• TO4222 development of the driver model
• development of the architecture
• calibration and validation of the model

3
Goal

• Provide an extensive model of drivers behavior
• integrating perceptive and cognitive information
  processing
• Producing a variety of behaviors, either car
  following or lane changing
• Assumption
• A driver has several objectives
• reach a destination
• maintain a satisfactory speed and/or gap with
  other vehicles
• Respect a time constraint

4
How to integrate constraints in the model?

• Development of the architecture of drivers
  cognition and perception
• perception was limited to scaling the relative
  velocity with a leading vehicle, the objective is
• to detect other elements that drivers use, such
  as traffic sign, direction sign, landmarks
• provide a better awareness of the surrounding
  traffic
• the cognitive system did not have the notion of
  goal, the objective is
• integrate the notion of goal across drivers
  activity
• process and preserve information
• Gather data for calibrating and supporting the
  development of the model

5
Perceptive System Development
6
Simulation Tool Development
7
Cognitive System Development

• Focus on drivers knowledge structure
• Strategic base of itineraries
• Tactical base of maneuvers
• Operational base of vehicle control
• Basis for processing this knowledge in regard to
  a current traffic situation
• Development of a mental representation
• Management of goals

8
Knowledge base
Operational
Traffic management
9
Transition

• Transition
• goal get to specific lane SpecLane which can
  be exit lane/split/merge

current lane / SpecLane
Maintain safe gap with leading vehicle
Determine number of lane to cross for reaching
SpecLane and set nb_of_lane to cross and id of
desired lane
if not available
Check availability of lane adjacent to current
lane
current lane lane adjacent toSpecLane
LaneId / desired lane id
if available
Send the command Initiate steering to operational
when seeing the rupture of the solid line
Move to adjacent lane and nb_of_lanenb_of_lane
1
When id of lane id of desired lane
10
Data collection

• Data protocol
• Drivers use the car for a 2 weeks period as they
  would use their personal car
• Once the driving test is completed, an interview
  is driven about morning and evening commute
• Data collection status
• 2 drivers out of 4 have performed the test
• 3rd driver is currently driving the vehicle

11
Instrumented Vehicle
12
Data reduction and processing for Driver 1
13
Morning commute description (1/2)
Landmark
Drive down on A
Look at
Direction
until you see 7-11
Driving
Make a right turn
(should be B) Pass a stop sign you will see C.
Go Straight, youll see Walgreens on the right,
continue straight until you see the Gas 76
station.
Make left.
After turn,
Stay on the far right lane for the ramp to 280 S
On the ramps stay on the left side, this will
take you to 280 S (not HWY1)
14
Morning commute description (2/2)
Landmark
Continue on 280
Look at
Direction
Exit 380 Interchange
Driving
to get on 101 S
Continue on 101
Until you see 3rd St in San Mateo Start moving to
the right lanes You want to take the 92 E towards
Hayward
Proceed on 92
To cross over the San Mateo Bridge After the
bridge stay on the left lane
15
Morning Commute
Origin
280 S
101 S
380 E
92 E
Destination
16
Afternoon Commute
101 N
92 W
17
Driving on 92 W
18
Radar Data (1/2)
19
Radar Data (2/2)
20
Parameter for driver model
21
Vehicle control
22
Concluding Remarks

• Progress on the simulation aspects of the model
• Number of perceptible clues, supporting more
  elaborated information processing
• development of knowledge aspects
• Data supporting
• Development of itinerary
• Development of typical maneuvers (exit, split)
• Vehicle control specifics for these maneuvers
• Integration of context
• Specification of the environment characteristics
  allowing variations on the production of a
  maneuver, due to either traffic or infrastructure
  constraints

23
Acknowledgement

• From INRETS
• Hélène Tattegrain-Veste, architecture
• Thierry Bellet, definition of architecture
• Aurélien Garcia, perception module
• From PATH
• Swe-Kuang Tan, implementation and adaptation of
  SmartAHS
• Joel VanderWerf, data processing
• Natasha Kourjanskaia, data coding software
• Lia Mandiro, video data coding
• The Experimental team, instrumentation of the
  Taurus
• All the volunteers who graciously participated to
  the data collection