1. Andrew H. Fagg: Symbiotic Computing Laboratory 1

2. Agent Control Points: How we communicate matters Using well-defined languages Discovery of languages Communicating in precise ways Augmenting languages: creating new concepts and abstractions Andrew H. Fagg: Symbiotic Computing Laboratory 2

3. Harold Set up: video conference between classroom and "robot". Robot is shown as torso only (dehumanize). Each side has a set of blocks. Instructor creates a simple model out of a set of blocks. Volunteer student attempts to command the robot to replicate the model. Andrew H. Fagg: Symbiotic Computing Laboratory 3

4. Our Robot "Primitives" are small elements: move up/down/left/right rotation open/close hand condition actions on sensors (force or tactile) color Andrew H. Fagg: Symbiotic Computing Laboratory 4

5. Our Robot II Commands are interpreted literally Non-understood commands are indicated with a gesture (shoulder shrug?) Actions should be largely consistent, but some noise should be modeled Andrew H. Fagg: Symbiotic Computing Laboratory 5

6. Our Robot III Students are encouraged to create higher- level concepts and actions: Pick up X, Put down X, Shape, Size Can define in terms of body state. For example, size can be defined by the distance between the two hands when they are on either side of the object Andrew H. Fagg: Symbiotic Computing Laboratory 6

7. Discussion Points What did the robot understand? actions and conditionals How reliable was the robot? What was frustrating in working with the robot? What worked well with the robot? What are good abstractions to build? Andrew H. Fagg: Symbiotic Computing Laboratory 7

8. Wrapping a Box Set up: every pair of students is issued a box, multiple tissue paper pieces, a roll of tape, paper and pencil Write down complete instructions for wrapping the box with the tissue paper Andrew H. Fagg: Symbiotic Computing Laboratory 8

9. How do we define success? Is the box completely covered? Is the wrapping neat? Do we minimize the amount of tape used? Is the box free of hands and the table? Andrew H. Fagg: Symbiotic Computing Laboratory 9

10. Testing Once instructions are complete, two pairs of students exchange instructions and attempt to follow them. –Successful? (possible) go back to refine instructions in pairs Group of four work together to define a new set of instructions. Present these instructions to another group of four Andrew H. Fagg: Symbiotic Computing Laboratory 10

11. Logic with Dominoes Dominoes exercises (A, B, C and D are inputs; X and Y are outputs) #1: NOT –A -> Y –B -> X #2: Combiner –A -> X –B -> X Andrew H. Fagg: Symbiotic Computing Laboratory 11

12. Logic Problems II #3: Splitter –A -> X –A -> Y #4: Blocker –B -> X –A, B -> nothing happens Andrew H. Fagg: Symbiotic Computing Laboratory 12

13. Logic Problems III AND/OR (circuit is the same, the assignment of true/false is different) –AC -> X –AD -> Y –BC -> Y –BD -> Y Andrew H. Fagg: Symbiotic Computing Laboratory 13

14. Logic Problems IV XOR –AC -> X –AD -> Y –BC -> Y –BD -> X Andrew H. Fagg: Symbiotic Computing Laboratory 14

15. Discussion Assigning true/false logic values to A/B/C/D/X/Y. We can choose this however we want NOT –A,Y=true; B,X=false Can choose the other way: same behavior Andrew H. Fagg: Symbiotic Computing Laboratory 15

16. Discussion II AND/OR –A/C/X = true; B/D/Y = false: AND –A/C/X = false; B/D/Y = true: OR –Demorgan's laws Computability: with NOT, AND, OR: we can compute anything a computer can compute Andrew H. Fagg: Symbiotic Computing Laboratory 16

17. Maps and Planning Treasure Island: demonstration with 3 islands What does the initial map look like? Interaction with the harbormaster Augment map with arrows Andrew H. Fagg: Symbiotic Computing Laboratory 17

18. Seven-Island Exercise Start with map that has just the islands shown. Each group/individual starts at Pirates' Island Indicate which boat to take: A or B (with no information from the harbormaster) Harbormaster directs this person to the appropriate island. Group notes this on their map Process continues. If stuck, reset to Pirate's Island. Goal: discover what all of the boats do. This will allow us to plan a shortest path from one island to another. Andrew H. Fagg: Symbiotic Computing Laboratory 18

19. Discussion What is the shortest path and the associated actions? Can describe the sequence of boats as a "plan". What is a path that will take 5 hops? What is a path that will take 6 hops? What is a path that will take 7 hops? Andrew H. Fagg: Symbiotic Computing Laboratory 19

20. Discussion II What happens if the boats cost doubloons? Example: all cost 1 and Pirate's Island B costs 2? What about costing 5? What if the boat captain will pay for you to travel from Shipwreck Bay to Dead Man's Island? 1? 5? Andrew H. Fagg: Symbiotic Computing Laboratory 20

21. Larger Discussion Contexts: path planning for cars; trucking; robot planning how to make a meal in your kitchen; phone trees Andrew H. Fagg: Symbiotic Computing Laboratory 21

22. Discussion III Reset all costs to 1. What if the B boat from Pirate's Island returned to Pirate's Island with probability p? Context: Game play; robots working in environments that they cannot entirely predict Andrew H. Fagg: Symbiotic Computing Laboratory 22