ARTIFICIAL INTELLIGENCE Summary Lecturer: Silja Renooij These - - PowerPoint PPT Presentation

ARTIFICIAL INTELLIGENCE

Lecturer: Silja Renooij

Summary

Utrecht University The Netherlands

These slides are part of the INFOB2KI Course Notes available from www.cs.uu.nl/docs/vakken/b2ki/schema.html

INFOB2KI 2019-2020

Subject overview

• Introduction (chapters 1,2)
• Pathfinding & Search (a.o. chapter 4)
• Learning (a.o. chapter 7)
• Deterministic planning & decision making

(a.o. chapter 5, 6, 8, 11)

• Reasoning and planning under uncertainty

(a.o. chapter 5)

• Movement (chapter 3)

2

What is Intelligence? What is (game) AI?

• Intelligence: A very general mental capability […] for

comprehending our surroundings—"catching on," "making sense" of things, or "figuring out" what to do

• Artificial Intelligence: four categories
• Game AI: is about the illusion of human behaviour

(Smart ‐‐ to a certain extent, unpredictable but rational, emotional influences, body language to communicate emotions, integrated in the environment) Thinking rationally Thinking humanly Acting rationally Acting humanly

3

Summary: Search & Pathplanning

• Used in environments that are static,

deterministic, observable, and completely known (accessible)

• Goal‐based
• Atomic states and actions (no domain structure)
• Uninformed search: BFS, DFS, UCS, IDS…
• Informed search: Greedy, A*,…
• Local search: hill‐climbing, simulated annealing,

GAs,…

• Adversarial search: minimax (alpha‐beta)

What about natural paths?

4

Natural paths: Short…(?)

Pre‐processing step:

• automatic construction of waypoint graph, using sampling

Automatic construction of roadmaps for path planning in games, Nieuwenhuisen, Kamphuis, Mooijekind & Overmars, 2004. 5

Natural paths: enough clearance (?)

Step 1: move waypoint c to ‘center’ cv between obstacles

Idea:

• find closest point on obstacle
• move in opposite direction until

same distance to other obstacle

6

Natural paths: enough clearance (?)

Step 2: move edges to ‘center’ between obstacles

Idea:

• Split edge with insufficient clearance
• move middelpoint to center (green line)

7

Natural paths: continuous (?)

• Add a circular blend to every pair of

incoming edges for every waypoint

• Curvature depending on the amount of

clearance

8

Natural paths result

9

Path planning: some problems remain…

• Characters will still all follow the same

path

• No dynamic evasion
• No natural behavior
• Characters take turns that have no

corresponding animation

• In‐game changes to the characters have

no effect on the path they follow

• Mood has no effect

(although claimed by some games, there is no real evidence of it’s use)

• …

Open problems

10

Summary: Learning

• Learning is essential for intelligence
• Learning is difficult when the world is

dynamic, large and uncertain

• Reinforcement learning does not necessarily

need a model of the world

• Supervised techniques learn from ‘examples’

– Decision trees, Naïve Bayes classifiers – Neural Networks; also used as function approximators

• Evolutionary algorithms used to let

strategies/solutions compete ~ search

11

Summary: Uncertainty

• How to reason with ‘uncertain’ information

– Fuzzy logic, Bayesian networks

• What is the best strategy if

– The outcome of actions is uncertain – The world state is uncertain

• MDP, POMDP used to determine optimal

actions (planning) under uncertainty

12

Summary Planning

Given initial state(s), goal(s) and set of actions: find a plan (a sequence of actions ) that is guaranteed to achieve goal(s).

• Inefficient as search: complete state descriptions,

too many actions, unused problem structure, multiple start and/or goal states, does plan exist?

• Different planning formalisms for problems with

different characteristics

13

Planning: problem characteristics

• Discrete (+ finite?) or continuous values
• Fully or partially observable
• One or more initial states
• Deterministic or stochastic
• With or without duration
• Concurrent or sequential
• Static or dynamic
• Reach goal state or maximize reward
• Single agent or multi‐agent

– Cooperative or selfish – Individual or centralized planning

States: Actions: Env.: Objective: Planners:

14

Classical: STRIPS (GOAP), NOAH, HTN (SHOP)

• Discrete (+ finite?) or continuous values
• Fully or partially observable
• 1 or more initial states
• Deterministic or stochastic
• With or without duration
• Concurrent or sequential
• Static or dynamic
• Reach goal state or maximize reward
• Single agent or multi‐agent

– Cooperative or selfish – Individual or centralized planning

D known 1 D W S S G S

Except GOAP! States: Actions: Env.: Objective: Planners:

15

Reactive: DT, FSM, BT, rule-based, …

• Discrete (+ finite?) or continuous values
• Fully or partially observable
• 1 or more initial states
• Deterministic or stochastic
• With or without duration
• Concurrent or sequential
• Static or dynamic
• Reach goal state or maximize reward
• Single agent or multi‐agent

– Cooperative or selfish – Individual or centralized planning

D S D S

States: Actions: Env.: Objective: Planners:

16

Stochastic: MDP, POMDP

• Discrete (+ finite?) or continuous values
• Fully or partially observable
• 1 or more initial states
• Deterministic or stochastic
• With or without duration
• Concurrent or sequential
• Static or dynamic
• Reach goal state or maximize reward
• Single agent or multi‐agent

– Cooperative or selfish – Individual or centralized planning

Often D ≥ 1 S W D R S MDP: F, POMDP: P

States: Actions: Env.: Objective: Planners:

17

Summary Moving

• Individual steering behaviors

– Intelligent animation? – What if an animation fails halfway?

• Group movement and teamwork

– Crowd simulation:

• Move together but separate
• Who stops first? Where do you stop? How do you

pass obstacles?

18

The MIRAnim Engine (for mixed reality)

Cassell et al. identify two types of communicative body motions:

• J. Cassell, T. Bickmore, M. Billinghurst, L. Campbell, K. Chang, Vilhjalmsson, H., and H. Yan. Embodiment in conversational interfaces: Rea. In Proceedings of

the CHI’99 Conference, pages 520–527, 1999.

Intelligent Animation

Performance Animation Interactive Virtual Humans

Gestures Posture shifts MIRAnim BLENDING

control idleness

19

Intelligent animation:

20

GRETA: Embodied Conversational Agent

Greta can talk and simultaneously show facial expressions, gestures, gaze, and head movements.

Intelligent Animation: problems

To make agent realistic/believable:

• Blend of facial animation and body

animation

• Blend of moving and handling objects
• Adjustment of moving/posture to

environment

• Animation shouldn’t get “stuck”
• Not too realistic?

21

Not covered in this course

• Constraint satisfaction
• Utility theory
• Natural Language Processing
• Vision/perception (recognizing objects)
• Knowledge representation (ontologies)
• Ethics
• …

22

Conclusions

• Much is achieved:

– Watson, Deep Blue – Alpha Go (Zero) – Robocup soccer – Autonomous cars – …

• With every step taken new challenges

become visible! (And not just in AI)

23

Interested in more?

• Search the internet 
• Course page
• Other courses ICS dept:

– Applied Games (Ba IKU) – Kennissystemen (Ba IKU) – Intelligente Systemen (Ba ICA) – Computationele Intelligentie (Ba ICA) – Probabilistic Reasoning (Ma COSC + AI) – Evolutionary Computing (Ma COSC + AI) – AI for game technology (Ma GMT) – …

• AI master (or COSC/GMT with electives)

Thank you!

(please use Caracal for feedback)

24