1. The Role of Intelligent Robotics in Teaching A.I. PACISE 2007 March 24 Mark M. Jones Oskars J. Rieksts

2. 2007Kutztown University2 Overview Motivation Motivation Reframing AI Reframing AI Main Areas of AI Main Areas of AI The Big 3 of Robotics The Big 3 of Robotics Intelligent Mobile Agents – Issues Intelligent Mobile Agents – Issues AI Issues in Robotics AI Issues in Robotics Learning from Experience Learning from Experience Conclusions Conclusions

3. 2007Kutztown University3 Motivation Fun – inherently interesting Fun – inherently interesting Attract students to CS Attract students to CS Attract students to AI Attract students to AI Undergrad research opportunities Undergrad research opportunities

4. 2007Kutztown University4 Reframing A.I. {the textbooks} Elaine Rich Elaine Rich  Artificial Intelligence  Search, GPS, STRIPS Russell & Norvig Russell & Norvig  A.I. : A Modern Approach  Intelligent agent Robin Murphy Robin Murphy  An Introduction to AI Robotics  Interactive applied A.I.

5. 2007Kutztown University5 The Main Areas of AI Knowledge Representation Knowledge Representation Inference and reasoning Inference and reasoning Learning Learning Planning Planning Natural language processing Natural language processing Search Search Vision Vision

6. 2007Kutztown University6 The Big 3 of Robotics Sense Sense Plan Plan Act Act

7. 2007Kutztown University7 Intelligent Mobile Agents - Issues Awareness Awareness Autonomy Autonomy Mobility Mobility Knowledge Knowledge Action Action Control Control Cooperation Cooperation Communication Communication

8. 2007Kutztown University8 Awareness Sensing Sensing  Situatedness  The world  Effect of actions  Self Knowing Knowing  What  Where  How  When

9. 2007Kutztown University9 Autonomy Purpose Purpose Direction Direction Decision Decision

10. 2007Kutztown University10 Mobility Location Location Change Change Direction Direction

11. 2007Kutztown University11 Action Purpose Purpose Planning Planning Decision Decision Behavior Behavior

12. 2007Kutztown University12 Control Coordination Coordination Timing Timing Knowledge Knowledge

13. 2007Kutztown University13 Cooperation Human-robot cooperation Human-robot cooperation Multi-robot cooperation Multi-robot cooperation Conscious coordination Conscious coordination Emergent behavior Emergent behavior

14. 2007Kutztown University14 Communication Human-robot communication Human-robot communication Robot-robot communication Robot-robot communication Sub-cognitive communication Sub-cognitive communication Cognitive communication Cognitive communication

15. 2007Kutztown University15 AI Issues in Robotics Spatial Cognition: A Case Study

16. 2007Kutztown University16 Spatial Cognition Definition: Definition:  Cognitive functions that enable people to deal effectively with spatial relations, visual spatial tasks and orientation of objects in space.  One aspect of these cognitive skills is spatial orientation, which is the ability to orient oneself in space relative to objects and events; and the awareness of self-location (A.S. Reber, The Penguin Dictionary of Psycology).

17. 2007Kutztown University17 Early History John Locke (1632 – 1704) John Locke (1632 – 1704)  Essay on Human Understanding  Tabula rasa = blank slate  Sensations strike blank mind  Mental processes produce vision of the world George Berkeley (1685 – 1753) George Berkeley (1685 – 1753)  Essay Toward a New Theory of Vision  Explored relationship between sight and touch..  and between ideas and the material world

18. 2007Kutztown University18 20 th Century George M. Stratton (1865-1957) George M. Stratton (1865-1957) Experiments with vision Experiments with vision  Inversion of visual input (special goggles)  Visual system builds anew  Gregory: “perhaps the most famous experiment in the whole of experimental psychology” The Spatial Harmony of Touch and Sight The Spatial Harmony of Touch and Sight

19. 2007Kutztown University19 Stratton’s Experiment Immediate consequence Immediate consequence  Perceptual-motor coordination disrupted  Disparity between seen and thought to be Restoration Restoration  Day 3 – “things interconnected into a whole”  Process – “piecing together parts of the ever-changing visual fields”  Visual system builds anew “growing from an isolated group of perceptions”  “correspondence, point by point, between touch and sight, is built up associationally” Conclusion Conclusion  by experience person learns “what visual position corresponds to any given tactual position”

20. 2007Kutztown University20 Cognitive Structuring Rudolph Carnap Rudolph Carnap  Der logische Aufbau der Welt  The Logical Structure of the World Aufbau Aufbau  building  setup  structure  superstructure Structured artifact Structured artifact

21. 2007Kutztown University21 Issues in Spatial Cognition Experiential Space Experiential Space  Ego-centric  Goal-oriented  Tight sensory-motor coupling  Minimal knowledge representation Rodney Brooks Rodney Brooks  Intelligence without representation Bickhard’s frog Bickhard’s frog Kirsh’s Earwig Kirsh’s Earwig

22. 2007Kutztown University22 Issues in Spatial Cognition Arkin, Murphy, et. al. Arkin, Murphy, et. al.  Cspace – configuration space  Path planning – plan, then act  A* search  Minimal conceptualization Problems Problems  Subgoal obsession  Lost opportunities Oliver Sacks Oliver Sacks  The Man Who Mistook His Wife for a Hat

23. 2007Kutztown University23 Issues in Spatial Cognition Benjamin Kuipers Benjamin Kuipers  Semantic Spatial Hierarchy » Artificial Intelligence, 1999 Aspects Aspects  Models human cognitive map  Multiple interacting representations  Both » Qualitative » Quantitative  Multiple levels

24. 2007Kutztown University24 Spatial Semantic Hierarchy - Levels Control level Control level  Local geometric maps Causal level Causal level  Behavior  States linked by actions Topological level Topological level  External ontology » Places » Paths » Regions  Explain observed causal level patterns

25. 2007Kutztown University25 Issues in Spatial Cognition - Navigation Foo, Warren, Tarr Foo, Warren, Tarr  Human spatial strategies Task : learn novel shortcut Task : learn novel shortcut  Metric cognitive map  Landmark-based navigation Conclusions Conclusions  Landmark-based predominates  Ants » 1 st use survey knowledge.. » then switch  People do not

26. 2007Kutztown University26 Spatial Cognition – Study Centers Transregional Collaborative Research Center Transregional Collaborative Research Center  Spatial Cognition: Reasoning, Action, Interaction  Established by the Deutsche Forschungsgemeinschaft (DFG) Deutsche Forschungsgemeinschaft (DFG)Deutsche Forschungsgemeinschaft (DFG)  Universities of Bremen and Freiburg Concerned with: Concerned with:  Knowledge about spatial environments » Acquisition » Organization » Utilization » Revision

27. 2007Kutztown University27 Spatial Cognition – Study Centers Research Issues Research Issues  Investigatin of human spatial cognition  Mobile robot navigation Goal Goal  Investigate cognitive foundations for human- centered spatial assistance systems.

28. 2007Kutztown University28 Spatial Cognition – Study Centers Spatial Cognition Priority Program Spatial Cognition Priority Program  Spatial behavior  Spatial inference  Spatial coding  Aspect maps  Ego space  Mental models  Robot navigation  Route knowledge/learning  Landmark usage

29. 2007Kutztown University29 AI Issues in Robotics Awareness Awareness  Vision  Sensor interpretation  Knowledge acquisition  Learning Autonomy Autonomy  Cognitive basis  Planning  Decision

30. 2007Kutztown University30 AI Issues in Robotics Mobility Mobility  Localization  Mapping »Spatial representation  Topological vs. metrical  Egocentric vs. allocentric  Semantic structure of space » Spatial reasoning » Experiential realization  Path Planning » Optimization

31. 2007Kutztown University31 AI Issues in Robotics Knowledge Knowledge  Knowledge representation  Knowledge acquisition  Learning  Metaknowledge Action Action  Planning  Decision  Understanding » Causality » Consequences

32. 2007Kutztown University32 AI Issues in Robotics Control Control  Control Paradigms » Supervisory » Reactive » Hybrid – deliberative/reactive » Role of autonomous control  Coordination » Sensing & acting » Timing

33. 2007Kutztown University33 AI Issues in Robotics Cooperation Cooperation  Awareness of others  Shared goals  Models of coordination  Supervisory  Teamwork  Unaware cooperation Communication Communication  Language » Understanding » Generation

34. 2007Kutztown University34 Sample Projects Simulated/Theoretical Mapping Mapping Path planning Path planning  Graph-based  Limo driver » Asymmetric TSP  Delivery robot Emergent behavior – flocking Emergent behavior – flocking Simulator construction Simulator construction

35. 2007Kutztown University35 Sample Projects – Real World From scratch From scratch Roomba-based Roomba-based  RooTooth  On-board processor {API} Create-based Create-based  Heat-seeking  Odor interpretation Rehabilitative assistant Rehabilitative assistant Robotics Tutorial - Robotut Robotics Tutorial - RobotutRobotut

36. 2007Kutztown University36 Learning from Experience Simulators Simulators  OS issues  Library incompatibilities  Poor documentation  Wasted time Hardware Hardware  Basic electronics  Component failure  Possible quagmire

37. 2007Kutztown University37 Key Lessons Maintain Maintain  Primacy of theoretical  Clear goal Find Find  Tested hardware  Good text Texts Texts  Murphy – AI Robotics  Arkin – Behavior-Based Robotics  Mataric – Create robot coursework

38. 2007Kutztown University38 Conclusions regarding Intelligent Robotics (IR)

39. 2007Kutztown University39 IR is AI-complete Key AI issues central to IR Key AI issues central to IR Examples discussed above Examples discussed above New Elsevier series – Foundations of Artificial Intelligence New Elsevier series – Foundations of Artificial Intelligence  Temporal reasoning  Knowledge representation  Autonomous systems and robots  Learning and vision  Spatial reasoning

40. 2007Kutztown University40 IR is CS-complete Behavior-based  OOP Behavior-based  OOP Interleaved control  OS Interleaved control  OS  Interrupts  Job scheduling Tractability  Algorithm D&S Tractability  Algorithm D&S Path planning  Search Path planning  Search Vision Vision Many other examples Many other examples