Scribd
Upload
3 views

M09 DeliberativeLevel

Uploaded by

Aaa Aa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
3 views

M09 DeliberativeLevel

Uploaded by

Aaa Aa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

High-level Decision-Making

Deliberative levels and computation


Deliberative robots: Introduction
• Autonomous robots act deliberately to fulfill their missions, i.e., they
perform actions to attain selected, reasoned objectives
• Deliberative levels
often follow
Belief-Desire-Intention (BDI)
architecture

https://learn.microsoft.com/en-us/archive/msdn-
magazine/2019/january/machine-learning-
leveraging-the-beliefs-desires-intentions-agent-
architecture

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 92
Deliberative robots: BDI architecture
1. The belief revision function (BRF)
updates the current beliefs
database, making sure that the
recently received set of percepts
doesn’t contradict any beliefs in
the belief dataset.
2. Desires are updated with new
options in regard of current
beliefs and intentions.
3. The set of desires are
transposed into intentions by
filtering unrealistic or hard-to-
fulfill ones.
4. Actions are selected from
intentions by following some
means-end reasoning approach.

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 93
Deliberative robots: BDI and state machines

BDI State machines


• BRF • Read inputs
• Beliefs • Current state
• Generate options • Future state computation
• Desires • Set of future states
• Filter • Select best future state
• Intentions • Next states (path to best future state)
• Actions • Write outputs

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 94
Deliberative robots: Architecture
• Deliberative layer
• Decision-making level
• Decides objectives (future states) and
fires plans to attain them
• E.g. Go to some spot
(for picking up/placing something)
• Executive layer
• Sequences operations to follow a plan
• E.g. Follow a path or
get a point-cloud from sensor
• Reactive layer
• Performs basic operations
• Interacts with the physical world
• E.g. Make the vehicle move
to a destination point

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 95
Deliberative robots: Making decisions
• Following pre-defined paths
• Select a path and send the corresponding command to the executive level
• Decide what to do in case command failed

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 96
Path planning
• State-machine model
• Like exercise 8 extension inside ReturnPath
• Programming model
ReturnPath = function(P)
local R = {"1 90 0"}
local a = -90
local d = 0
local i = #P
while i>=0 do
d = tonumber(string.match(P[i] or "", "%d+[^%d]*-?%d+[^%d]*(%d+)")) or 0
if a~=0 or d~=0 then table.insert(R, string.format("1 %d %d", -a, d)) end
a = tonumber(string.match(P[i] or "", "%d+[^%d]*(-?%d+)")) or 0
i = i - 1
end -- while
table.insert(R, "1 90 0")
table.insert(R, "1 90 0")
return R
end --- ReturnPath

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 97
Deliberative robots: Making decisions
• L2 level: User interaction

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 98
Deliberative robots: Making decisions
• Example case: Travelling predefined paths with alternatives
• In case of failure of following path P,
return to origin and try alternative path P2
• In case of blockage when travelling return or alternative path,
stop operation
• Use function calls to get data

© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 99
Deliberative robots: Making decisions
© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona

106589 Intelligent Robots, 100


Exercise 9: Implement L2 EFSM
• Functions sub() and rP() are given
• Interactive, stand-alone version requires users enter data from…
• … the user interface
> Navigator.button (0: QUIT 1: NAVIGATE, 2: START, 3: SWEEP else none) = ?
• … and the L1 layer
> Navigator.A (answer from L1) = ?
• You may type in
• “D” for done;
• “E” for error;
• “E 3 D BLOCKED 15” for blocking situations that have timeout
at 15 cm of end point of step no. 3 of path, and the like
• Use M09_L2_Navigator.lua
• Main and alternative paths as parameters of Navigator.init()
© 2023 Lluís Ribas i Xirgo, Universitat Autònoma de Barcelona 106589 Intelligent Robots, 101

You might also like