Conducting Situated Learning in a Collaborative Virtual Environment

1
Conducting Situated Learning in a Collaborative
Virtual Environment

• Yongwu Miao
• Niels Pinkwart
• Ulrich Hoppe

2
Overview

• Pedagogical motivation constructivism and
  situated learning
• Approach and principles of 3D collaborative
  driving simulator
• Implementation key decisions (driving place,
  situation detection, architecture for
  distribution)
• Example scenes
• Conclusions and future work

3
3D Simulations as constructivist learning
environments

• Core position of constructivism learners
  actively construct knowledge
• Knowledge based on interpretation of experiences
  in the real world (includes other learners!)
• 3D Simulations of real world sometimes very
  appropriate (costs, safety) learners can still
  be active and make experiences
• Example learning car driving

4
Existing systems

• Lots of 3D car driving simulators exist (games,
  educational, professional)
• Educational systems typically try to confront
  learners with challenging situations
• Often full size systems very costly (advanced
  visual and audio systems, motion systems,
  functional cab, software components)
• Growing PC and network performance allows low
  cost solutions usually with pre-defined
  driving scenarios and tutors

5
Our approach

• Low cost (standard PC and network), support for
  multiple users
• Variety of challenging situations that might
  happen through interaction / collaboration no
  predefined scenes!
• Consider situated learning principles

Content
Community
Learner
Context
Participation
6
Driving place design

• Key requirement rich data model (realistic
  content context), but still small enough for
  distributed usage
• General approach cell grid
• Each cell containing typed objects (static or
  dynamic) with attributes
• Example car object with attributes direction,
  speed, acceleration, turning angle, brake status,
  indicator status, sector information

7
Map editor

• Create driving places easily by drag drop
• Maps transformed to VRML
• Display via Java 3D

8
Situation description and recognition

• Not needed for most basic functionality (except
  collision detection)
• Essential for advanced functions (user behavior
  analysis, feedback)
• Technical approach Jess rules acting on object
  attributes
• Situation detection ? target specification
• Additional control rules to check if targets have
  been reached

9
Example situation recognition

• (defrule safe_distance_violation
• (vpcar (position ?pos)
• (direction ?dir) (speed ?speed))
• (car_in_lane
• (car_position ?carpos) (car_direction
  ?cardir)
• (car_speed ?carspeed))
• (not (target_state (desc safe_distance_violation)
  ))
• (test (violated_safe_distance ?pos ?speed
  ?carpos ?carspeed))
• gt
• (bind ?list (create "distance"))
• (?guidance addInstruction 6 ?carpos ?list
  ?pos)
• (assert (target_state
• (situid 6)
• (checkpoint ?carpos)
• (chkpt_passed FALSE)
• (targets ?list)
• (desc safe_distance_violation)))
• (?guidance addMistakes ?list 6))

Attributes of students car
Attributes of other car in lane
Distancetoo small ?
Definition of new target
10
Distributed system architecture

• Central tuple space contains attributed objects
  (driving place and additional information)
• Different roles for teacher and student client

11
Distributed system architecture

• Reduction of network traffic
• Transmission of only local context (sector
  arithmetic)
• Only status change events (braking,accelerating,
  indicator) for cars, positions are inferred by
  clientapplications

12
Feedback

• Based on situations recognition and targets,
  different types of feedback and guidance
  possible
• Forewarn messages or hints
• Feedback after targets missed/reached
• Implicit feedback (situation creation)
• Guidance on demand

Alreadyimplemented
13
System architecture

• System prototype (simple graphics, small number
  of object types, restricted number of modeled
  situations) exists and has been used in a pilot
  study

14
Example student client
15
Example teacher client
16
Conclusions

• Low-cost collaborative 3D educational driving
  simulator, following situated learning approach
• Allows training in a lot (though not all) of the
  skills needed for driving
• No hard-coded challenging situations created by
  system, but (more realistic!) provision for
  collaborative situation creation
• Students receive feedback on their performance in
  real-time

17
Future Work

• Agents simulating students
• Subtle creation of situations by intelligent
  agents
• Integration of audio communication functions
• Evaluations beyond pilot tests
• Email nielsp_at_cs.cmu.edu