Section 3 : Conditional, continuous, and multiagent planning

Commentary

Section Goals

• To discuss conditional planning problems, and the relevant algorithms in both fully or partially observable environments.
• To introduce continuous planning and multiagent planning problems, and the basic idea behind them.

Learning Objectives

Learning Objective 1

• Explain how conditional choices are performed in planning (conditional planning), and what algorithms fit the problems.
• Outline the and-or-graph-search algorithm for conditional planning in fully observable environments.
• Explain how belief states are represented in partially observable environments for conditional planning.
• Outline the basic principle of continuous planning.
• Outline the main issues and problems that are handled in multiagent planning, such as joint goals, plans, actions, and communications.
• Explain the coordination mechanisms in multiagent planning, and some special outcomes, such as emergent behaviour and joint intention.
• Explain the following concepts or terms:
  • Conditional effects
  • And-Or graph
  • Belief state
  • Online planning
  • Joint goal
  • Joint plan
  • Joint action
  • Convention and social law
  • Emergent behaviour

Objective Readings

Required readings:

Reading topics:

Nondeterministic and MultiAgent Planning (see Sections 11.3 - 11.4 of AIMA3ed).

Objective Questions

• Why are belief state representation and maintenance important for planning in partially observable environments?
• What are the possible applications of conditional planning and multiagent planning?
• What are the benefits and challenges of AI planning, such as nondeterministic planning and multiagent planning, when compared to non-AI methods?

Objective Activities

• Analyse the following algorithms introduced in the textbook about planning.
  • Hierarchical-Search
  • Angelic-Search
• Complete Exercise 11.10 of AIMA3ed.
• Complete Exercise 11.12 of AIMA3ed.