SmartAHS: Concept and Tools

1
SmartAHS Concept and Tools

• Presented by Raja Sengupta
• Principal Researcher, PATH LAB
• California PATH, UC Berkeley
• Co-workers Many people at PATH, EECS, and ME
• Sponsors NAHSC, CALTRANS, ONR, ARO

2
Transportation System Modeling is a Hard Problem
Complex, Large-Scale System of Systems

• Subsystem interactions are complex
• ITS is a technology revolution
• Traffic dynamics caused by human behavior
• Even more complex!

3
Common Components of the Microscopic
Transportation Simulator
4
There are many Valid combinations
5
There are many Valid combinations
6
Each tool developer invests in a particular
combination of models

• The right combination for a system depends on
  many factors
• Modeling objectives
• Throughput / Safety / Emissions / Travel Demand ?
• Computational resources
• PC, Workstation (s), or Cray?
• Available modeling data
• Count data, toll collection data, regional
  planning model data
• Deployed technologies
• TMCs, signal control systems, information
  services
• So the Tool Developer has a hard problem
• Our research objective
• Techniques for tailorable simulators
• Benefits both developers and users

7
Tailorable Simulators The SHIFT Approach

• SHIFT is a language for writing simulations
• Use the language to write libraries of component
  types
• Use the libraries to tailor a new simulator for
  each application
• SmartAHS, MOB SHIFT, SmartUCAV, California
  SmartBRT, GM Threat Assessment Simulator
• To tailor a simulator
• Choose models from each library
• For the few (hopefully!) new technologies
• Sub-type a model in an existing library
• Create a new type or library
• Write SHIFT Glue
• Compile, link, and play

8
Example Assembling a really Micro-Simulator
9
Put the Modeling Technology in LibrariesC,
Matlab, Lisp, SHIFT
Roadway Model Library
Driver Model Library
Vehicle Model Library
Control Model Library
Sensor Model Library
Signal Control Library
10
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Roadway Model Library

• I-15 San Diego cartographic model
• Houston Katy graph model

Driver Model Library
Vehicle Model Library
Control Model Library
Sensor Model Library
Signal Control Library
11
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Roadway Model Library

• I-15 San Diego model
• Houston Katy graph model

Driver Model Library
Vehicle Model Library
Control Model Library
Sensor Model Library
Signal Control Library
12
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library

• STS Car following model
• Driver with ACC model
• Cognitive Driver Model

Vehicle Model Library
Signal Control Library
13
General Architecture of the Cognitive Driver Model
R O A D E N V I R O N M E N T
Legend Writing/reading Parameters Message
Strategic
Tactical
Perception
Management and control
Operational
Module
Execution
Emergency management
COSMODRIVE (BELLET 1998)
14
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library

• STS Car following model
• Driver with ACC model
• Cognitive Driver Model

Vehicle Model Library
Signal Control Library
15
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library
Vehicle Model Library

• 1-d Particle model
• 2-d Particle model
• 2-d Dynamic model
• 6-dof Dynamic model
• 4 sprung mass

Signal Control Library
16
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library
Vehicle Model Library

• 1-d Particle model
• 2-d Particle model
• 2-d Dynamic model
• 6-dof Dynamic model
• 4 sprung mass

Signal Control Library
17
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library
Vehicle Model Library

• ACC without braking
• ACC with braking
• CMU Autonomous Driving
• Platoon driving

Signal Control Library
18
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library
Vehicle Model Library

• ACC without braking
• ACC with braking
• CMU Autonomous Driving
• Platoon driving

Signal Control Library
19
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library
Vehicle Model Library

• Perfect sensor model
• GM Swell radar model

Signal Control Library
20
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Driver Model Library
Vehicle Model Library

• Perfect sensor model
• Gm Swell radar model

Control Model Library
Sensor Model Library
Signal Control Library
21
Example Assembling a really Micro-Simulator
Step 1 - Pick your models
Roadway Model Library
Driver Model Library
Vehicle Model Library
Control Model Library
Sensor Model Library
22
Example Assembling a really Micro-SimulatorStep
1 - Pick your models
Roadway Model Library
Driver/Vehicle Type
Driver Model Library
Write SHIFT Glue
Vehicle Model Library
Control Model Library
Sensor Model Library
Traffic Source Library
23
Example Assembling a really Micro-SimulatorStep
2 - Glue your models
Driver/ Vehicle Type
Roadway
Traffic Source Type
24
Example Assembling a really Micro-SimulatorStep
2 - Glue your models
Final SHIFT Simulator Program
25
Graphical User Interface
26
SHIFT Software Process
SHIFT Program
Run-time Executable
Run-time Output
Parser/Type Checker
MS Windows/ UNIX
IR
Run-time Libraries
Linker
Animator
Run-time Executable
Movie
27
The really Micro-simulatorACC Case Study

• Investigating the effects of ACC in stop-and-go
  traffic using SmartAHS
• real I-880 traffic data
• assume no cut-ins, i.e., single lane
• ACC control laws used in the UMTRI field
  operational test
• detailed vehicle models
• Examining mobility, utility, safety

28
Assembling a less Micro-Simulator
29
The Houston Katy-Freeway AHS Case StudyPicking
the Models
Roadway Model Library

• Houston Katy graph model

Driver Model Library

• 2-d Dynamic model

Vehicle Model Library

• CMU Autonomous Driving
• Platoon driving

Dynamic ODs based on Houston Metro data
Demand Model Library
30
The Houston Katy-Freeway AHS Case StudyAssemble
and Run
Road
Vehicle
Control
Sensor
Demand
31
The Houston Katy-Freeway AHS Case
StudyAssemble, Compile, and Run
Assemble
Road
Vehicle
Compile
Control
Sensor
Run
Demand
32
The Houston Katy-Freeway AHS Case Study
33
How does the same tool produce simulations of
widely different granularity?

• Tailorability
• The SHIFT programming language
• SHIFT simulations consist of SHIFT components
• SHIFT is loosely object-oriented
• Components are created from types
• A generic source in the demand library

type source state number highway_section_id
timer 0 vehicle car flow timer
1 ... when timer 1 do timer 0 car
create (vehicle , speed rand(20,30))
34
The SHIFT Programming LanguageInheritance

• Inheritance facilitates plug and play
• If the complex vehicle is a subtype of vehicle
  the source type is reusable

type complex_vehicle vehicle
type source . . when timer 1 do timer
0 car create (complex_vehicle , speed
rand (20,30))
35
The SHIFT Programming LanguageInterfacing
Components

• Components interface through
• continuous data feeds
• messages
• type name input ... //variables written
  by others output ... //variables read by
  others state ... //internal variables

36
The SHIFT Programming LanguageInterfacing
Continuous Data Feeds
throttle
throttle
Vehicle dynamics
controller
Car

• type car state Controller ctrl
  VehicleDynamics dyn setup connect
  throttle(dyn) lt- throttle(ctrl)

37
The SHIFT Programming LanguageInterfacing
Components through Messages
My_Car
Your_Car

• Type my_car car ... export following
  transition free-gtfollowfollowing
• type your_carcar ...
• state car behind_car
• transition dummy -gt dummy behind_carfollowing

38
Summary

• Current Tools
• Libraries
• Vehicle, Driver, Sensor, Communication, Control
• Two road networks Houston Katy, I-15 San Diego
• GUI
• Graphical Highway Builder
• Translator, Run-time libraries
• Past Use
• SmartAHS, MOB-SHIFT, SmartUCAV
• Daimler-Chrysler for Inter-Vehicle Communications
  
• Faculty of Engineering, Porto, Underwater
  Autonomous Vehicles

39
Conclusions The Open Model for Simulator
Development

• A simulator is a package created from libraries
• Object-oriented architecture
• Clear component interfaces
• Open architecture
• Public interfaces
• Private implementations
• Technology of the vendor would be in the
  libraries and in the packages
• Future Work
• Simulation Technologies
• Adaptive Simulation
• Concurrent Multi-scale simulation
• Applications to ITS
• SmartBRT, GM Threat Assessment Tool

40
Concurrent Multi-Scale Simulation
Large simulations on small machines by simulating
at multiple fidelities
Detailed Study Area
41
Adaptive Simulation

• Standard object-oriented (oo) paradigms usually
  have 1 level of abstraction there is a class
  and instances of it
• Metaobject protocols abstract classes by making
  them instances of some other class I.e. the
  metaclass
• Since classes and objects determine the behaviour
  of your language metaclasses act as
  specializers to extend the semantics of your
  language
• Components can now create classes too!
• Use the simulations to improve the simulator

42
SHIFT and LISP

• Why LISP? Expressiveness, utility libraries
  (e.g. xlispstat, CLASP, CLIM, CL-HTTP, ) for
  mathematics, GUI, web applications...
• Interpreter allows for rapid development and
  immediate testing, debugging capabilities.

43
Metaclasses and SHIFT

• Weve implemented SHIFT with the MOP for CLOS
  (Common Lisp Object System) so that we may unify
  a different set of protocols (I.e. semantics) to
  the same SHIFT models.

STANDARD HYBRID CLASS
CAUSE-EFFECT/ ZERO-ORDER HOLD MODEL
SYNCHRONOUS/RK-4 MODEL
ENVIRONMENT
CAR
CAR 2
CAR 1
ROAD 1
44
The SHIFT Programming LanguagePowerful
Constructs for Concise Programming

• General set operations
• transition start -gt free
• free -gt follow when exists k in cars abs(x(k)
  - x) lt 50 do front_vehicle k ,
• Mode-Switched Differential equations
• state
• free free_flow, follow follow_flow
• flow
• free_flow xv, va, a F(65) ,
• follow_flow xv, va, a F(v(front_vehicle)