RadarSpaceTime

1
RADAR/Space-Time AssistantCurrent and Future
Research
2
Current system
Representation
Optimizer
Elicitor
3
Current system
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI

• Current
• Scheduling of rooms

4
Current system
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI

• Current
• Scheduling of rooms
• Basic defaults

5
Current system
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI

• Current
• Scheduling of rooms
• Basic defaults
• Elicitation questions

6
Current system
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI

• Current
• Scheduling of rooms
• Basic defaults
• Elicitation questions
• Almost no bugs

7
Near-future system
Top-level control
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI

• Current
• Scheduling of rooms
• Basic defaults
• Elicitation questions
• Almost no bugs

8
Near-future system
Top-level control
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI
Console
RADAR

• Current
• Scheduling of rooms
• Basic defaults
• Elicitation questions
• Almost no bugs

• Near-future
• Choice of rooms
• Scheduling of equipment

9
Near-future system
Top-level control
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI
Console
RADAR

• Current
• Scheduling of rooms
• Basic defaults
• Elicitation questions
• Almost no bugs

• Near-future
• Choice of rooms
• Scheduling of equipment
• Intelligent defaults

10
Near-future system
Top-level control
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI
Console
RADAR

• Current
• Scheduling of rooms
• Basic defaults
• Elicitation questions
• Almost no bugs

• Near-future
• Choice of rooms
• Scheduling of equipment
• Intelligent defaults
• Integration with RADAR

11
Near-future system
Top-level control
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI
Console
RADAR

• Short-term research
• Learning of default preferences

12
Near-future system
Top-level control
backend
Representation
Optimizer
Elicitor
Plumbing
frontend
GUI
Console
RADAR

• Short-term research
• Learning of default preferences
• Learning of elicitation questions

13
Long-term research

• Contingency scheduling
• Learning to collaborate with the user
• Opportunistic and transfer learning
• Planning and learning of strategies

14
Contingency scheduling
Top-level control
Representation
Optimizer
Elicitor
Plumbing
GUI
Console
RADAR

• Identify critical uncertainties
• Generate alternative schedules for different
  possible scenarios
• Maximize similarity among these schedules
• Help the user keep track of the alternatives

15
Collaboration learning
Top-level control
Representation
Optimizer
Elicitor
Plumbing
GUI
Console
RADAR

• Collect data on the users decisions at
  different steps of scheduling
• Assign credit or blame to each decision
• Identify the areas of the users strengths,
  as compared to automated decisions

16
Opportunistic and transfer learning
Top-level control
Representation
Optimizer
Elicitor
Plumbing
GUI
Console
RADAR

• Collect all useful information
• Identify learning opportunities
• Transfer the learning results among tasks
  within Space-Time Assistant
• Make them available to other modules

17
Strategy planning and learning
Top-level control
Representation
Optimizer
Elicitor
Plumbing
GUI
Console
RADAR

• Determine possible sequences of high-level
  actions
• Search for most effective sequences
• Learn heuristics and control rules that
  improve the search efficiency