The BDI control loop

Technique 1: the basic loop, cycle (1)

Problem it solves: a first runnable version of BDI.

1. B := B0;                   /* initial beliefs */
2. while true do
3.     get next percept ρ;
4.     B := brf(B, ρ);        /* belief revision */
5.     I := deliberate(B);
6.     π := plan(B, I);
7.     execute(π)
8. end-while

The flaw (slide 6): if deliberation starts at \(t_0\) and the plan is executed up to \(t_2\), the plan was right for the world at \(t_0\). That’s calculative rationality.

Technique 2: deliberation = options + filter, cycle (2)

Problem it solves: separates “what could I do?” from “what will I commit to?”.

\(options\) takes the current intentions as input: what you’ve already committed to shapes what you’ll consider (“intentions filter options”).

1. B := B0;  2. I := I0;
3. while true do
4.     get next percept ρ;    5. B := brf(B, ρ);
6.     D := options(B, I);    7. I := filter(B, D, I);
8.     π := plan(B, I);       9. execute(π)
10. end-while

   lab ─── hall ─── office
    │                 │
  garden ─── yard ────┘

Technique 3: commitment strategies

Problem it solves: “how long should an intention last?”

The book’s story (p. 76): Willie the household robot. Asked for a beer, the first Willie does something else (under-committed). The fixed Willie fetches the beer even after being told “never mind” (over-committed). The final Willie smashes the last bottle on the way: its rules let it drop commitments that became impossible, and it made this one impossible itself.

Commitment applies to two things (slide 10): the ends (the intention) and the means (the plan). Cycle (2) is over-committed to both.

Technique 4: commitment to means, cycle (3)

Problem it solves: notices a broken plan and fixes it, without changing the goal.

Execute one action at a time and look after each:

9.  while not empty(π) do
10.     α := head(π);  11. execute(α);  12. π := tail(π);
13.     get next percept ρ;  14. B := brf(B, ρ);
15.     if not sound(π, I, B) then
16.         π := plan(B, I);
17.     end-if
18. end-while

Technique 5: commitment to ends, cycle (4)

Problem it solves: stops working on a goal that’s already done or can never be done.

Cycle (4) changes only line 9:

9.  while not (empty(π) or succeeded(I, B) or impossible(I, B)) do

What to read in the book

Practice prompt

I'm studying CS5940 Intelligent Agents (Wooldridge, An Introduction to MultiAgent Systems, 2nd ed., 4.3). I want practice on THE BDI CONTROL LOOP, cycles (1) to (4):
- the basic loop (perceive, brf, deliberate, plan, execute) and why it suffers from calculative rationality
- the types and roles of brf, options, filter and plan, and why options takes the current intentions
- blind, single-minded and open-minded commitment; the Willie robot story
- commitment to means (cycle 3: execute one step, perceive, replan if not sound) vs commitment to ends (cycle 4: stop on succeeded or impossible)
- tracing a robot through a small map of rooms when a door closes or someone else achieves the goal: count actions and wasted actions for each cycle
- why the slides' labels for cycles (2) and (3) are a little loose

Please:
1. Use plain, everyday language and explain any symbol in one short sentence.
2. Give me ONE question at a time, easy first, then harder; describe maps and events precisely.
3. Wait for my answer, then mark it and show the full trace or model answer.
4. If I get it wrong, explain the mistake and give me a similar question.
5. Every 5 questions, summarise my strengths and what to practise.