Intelligent Robotics

MAP-I Doctoral Programme in Computer Science

Luís Paulo Reis (University of Porto)

Nuno Lau (University of Aveiro)
http://paginas.fe.up.pt/~lpreis/ir2007_08/

Questions?
• Questions about Introduction?
• Questions about Sensors and Perception?
• Questions about Ciber-Mouse Simulator?
• Any questions about Assignment #1?

2
 “A robot in every home”
• In the January 2007 issue of Scientific American Bill
Gates wrote:
– As I look at the trends that are now starting to converge, I can
envision a future in which robotic devices will become a nearly
ubiquitous part of our day-to-day lives
– The challenges facing the robotics industry are similar to those
we tackled in computing three decades ago
– Robotics companies have no standard operating software that
could allow popular application programs to run in a variety of
devices
– The standardization of robotic processors and other hardware is
limited, and very little of the programming code used in one
machine can be applied to another
– The goal of Microsoft Robotic Studio is to deliver a set of
programming tools that would provide the essential plumbing so
that anybody interested in robots with even the most basic
understanding of computer programming could easily write
robotic applications that would work with different kinds of 3
hardware

Some Robotic Issues

• Agent/Robot control architectures
• Behavior-based systems
• Sensors and Perception
• Representation Issues
• Adaptation and Learning
• Path planning and Navigation
• Localization and Mapping
• Intelligent Planning
• Multi-robot systems

4
 Some Robotic Issues
• How do I interpret my sensor feedback to determine my
current state and surroundings? [sensor
processing/perception]
• Where am I? [localization]
• How do I make sense of noisy sensor readings?
[uncertainty management]
• How do I fuse information from multiple sensors to
improve my estimate of the current situation? [sensor
fusion]
• What assumptions should I make about my
surroundings? [structured/unstructured environments]
• How do I know what to pay attention to? [focus-of-
attention]

Some Robotic Issues

• What should my control strategy be to ensure that I
respond quickly enough? [control architecture]
• How should I make decisions? [reasoning, task
arbitration]
• Where do I want to be, and how do I get there? [path
planning, navigation]
• I have lots of choices of actions to take --what should I
do in my current situation? [action selection]
• How should I change over time to respond to a dynamic
environment? [learning, adaptation]
• Why doesn’t the same action that worked in this situation
beforenot work now? [hidden state]
• How can I change my state from A to B [planning]
• How should I work with other robots? [multi-robot
cooperation, communication] 6
 Robotic Architecture - definition(s)
• An architecture provides a principled way of
organizing a control system. However, in
addition to providing structure, it imposes
constraints on the way the control problem can
be solved [Mataric, 1992]
• Robotic architecture usually refers to software,
rather than hardware [Arkin, 1998]
• How the job of generating actions from percepts
is organized [Russel and Norvig, 2002]

Spectrum of Robot Control

Architectures
• Deliberative control: “think hard, then act”
• Reactive control: “don't think, (re)act”
• Hybrid control: “think and act in parallel”

Model-Based Behavior-Based

Adapted from Arkin, Behavior-based Robotics (MIT Press, 1998)

8
 Typical Organizations
• Typical organizations:
– Hierarchical
– Behavior-based / Reactive
– Hybrid
• Reactive
– Cheap low memory processing
– No world model
• Deliberative
– Making maps
– Selecting behaviors
– Monitor performance
– Planning
• Hybrid deliberative/reactive paradigm

Deliberative Architectures
• In Deliberative control, the robot takes all of the
available sensory information, and all of the internally
stored knowledge it has, and it thinks ("reasons") about it
in order to create a plan of action.
• To do so, the robot must search through potentially all
possible plans until it finds one that will do the job. This
requires the robot to look ahead, and think in terms of: "if
I do this next, and then this happens, then what if I this
next, then this happens,..." and so on.
• This can take a long time, which is why if the robot must
react quickly, it may not be practical. However, if there is
time, this allows the robot to act strategically.

10
 Reactive Architectures
• Reactive Control is a technique for tightly coupling
sensory inputs and effector outputs, to allow the robot to
respond very quickly to changing and unstructured
environments.
• Think of reactive control as as "stimulus-response". This
is a powerful control method: many animals are largely
reactive.
• Limitations to this approach are that such robots,
because they only look up actions for any sensory input,
do not usually keep much information around, have no
memory, no internal representations of the world around
them, and no ability to learn over time.

11

Hybrid Architectures
• In Hybrid Control, the goal is to combine the
best of both Reactive and Deliberative control.
In it, one part of the robot's "brain" plans, while
another deals with immediate reaction, such as
avoiding obstacles and staying on the road.
• The challenge of this approach is bringing the
two parts of the brain together, and allowing
them to talk to each other, and resolve conflicts
between the two.
• This requires a "third" part of the robot brain,
and as a result these systems are often called
"three-layer systems“
Adapted from http://www-robotics.usc.edu/~maja/robot-control.html

12
 Deliberative vs Reactive
• No single approach is "the best" for all robots; each has
its strengths and weaknesses
• Control requires some unavoidable trade-offs because:
– Thinking is slow
– Reaction must be fast
– Thinking allows looking ahead (planning) to avoid bad actions
– Thinking too long can be dangerous (e.g., falling off a cliff)
– To think, the robot needs (a lot of) accurate information
– The world keeps changing as the robot is thinking, so the slower
it thinks, the more inaccurate its solutions
• As a result of these trade-offs, some robots don't think at
all, while others mostly think and act very little.
– It all depends on the robot's task and its environment!
13

Model-based - Deliberative
• Sense-plan-act paradigm: dominant view in the AI community was
that a control system for an autonomous mobile robot should be
decomposed into three functional elements [Nilsson, 1980]:
– a sensing system (translate raw sensor
input into a world model,
– a planning system (take the world model and
a goal and generate a plan to achieve the goal)
– and an execution system (take the plan and
generate the actions it prescribes)
• Perception is the establishment and
maintenance of correspondence between
the internal world model and the external
real world [Albus 1991].
• Action results from reasoning over the world
model.
• Perception is not tied directly to action.

14
 Reactive
• General assumptions:
– The environment lacks temporal consistency
and stability
– The robot’s immediate sensing is adequate
for the task at hand
– It is difficult to localize a robot relative to a
world model
– Symbolic representational world knowledge is
of little or no value
15

Behavior-Based - Reactive
• Common features:
– Emphasis on the importance of coupling sensing and action
tightly.
– Avoidance of representational symbolic knowledge (because the
world can change over time and uncertainty is hard to model).
– Decomposition into contextually meaningful units (behaviors or
situation-action pairs).
• Distinctions:
– Granularity of behavioral decomposition
– Basis for behavior specification (ethological, situated activity, or
experimental)
– Response encoding (e.g., discrete or continuous)
– Coordination methods (e.g., competitive vs cooperative)
– Programming methods, language, reusability

16
 Hybrid Architectures
• Combine the responsiveness, robustness, and flexibility of purely
reactive systems with more traditional symbolic/deliberative methods
• Reason: purely reactive systems lack the ability to take into account
a priori knowledge (e.g. about the world) and to keep track of the
history (memory)
• Typical three-layer hybrid architecture
– Bottom layer is the reactive/behavior-based layer,
in which sensors/actuators are closely coupled
– Upper layer provides the deliberative component
(e.g., planning, localization)
– The intermediate between the two is sometimes
called supervisory layer
• Examples of coupling between planning
and reactive layers:
– Planning to guide reaction: planning sets
reactive system parameters.
– Coupled: planning and reacting are
concurrent activities, each guiding
the other 17

Subsumption Architecture [Brooks 1986]

• Behaviors are augmented finite state machines (AFSM)
• Stimulus or response signals can be suppressed or inhibited by
other active behaviors; a reset input returns the behavior to its start
conditions
• Each behavior is responsible for its own perception of the world
• Arrangement in layers: lower layers have no awareness of higher
layers

18
 Brooks – Behavior languages
Brooks has put forward three theses:
1. Intelligent behavior can be generated without
explicit representations of the kind that
symbolic AI proposes
2. Intelligent behavior can be generated without
explicit abstract reasoning of the kind that
symbolic AI proposes
3. Intelligence is an emergent property of
certain complex systems

19

Brooks – Behavior languages

He identifies two key ideas that have informed his
research:
1. Situatedness and embodiment: ‘Real’ intelligence is
situated in the world, not in disembodied systems
such as theorem provers or expert systems
2. Intelligence and emergence: ‘Intelligent’ behavior
arises as a result of an agent’s interaction with its
environment. Also, intelligence is ‘in the eye of the
beholder’; it is not an innate, isolated property

20
 Brooks – behavior languages
• To illustrate his ideas, Brooks built some based on his
subsumption architecture
• A subsumption architecture is a hierarchy of task-
accomplishing behaviors
• Each behavior is a rather simple rule-like structure
• Each behavior ‘competes’ with others to exercise
control over the agent
• Lower layers represent more primitive kinds of
behavior (such as avoiding obstacles), and have
precedence over layers further up the hierarchy
• The resulting systems are, in terms of the amount of
computation they do, extremely simple
• Some of the robots do tasks that would be impressive if
they were accomplished by symbolic AI systems

21

A Traditional Decomposition of a Mobile Robot

Control System into Functional Modules

From Brooks, “A Robust Layered Control System for a Mobile Robot”, 1985

22
A Decomposition of a Mobile Robot Control
System Based on Task Achieving Behaviors

From Brooks, “A Robust Layered Control System for a Mobile Robot”, 198523

Layered Control in the Subsumption

Architecture

From Brooks, “A Robust Layered Control System for a Mobile Robot”, 198524
 Example of a Module – Avoid

From Brooks, “A Robust Layered Control System for a Mobile Robot”, 198525

Schematic of a Module

From Brooks, “A Robust Layered Control System for a Mobile Robot”, 198526
 Levels 0, 1, and 2 Control
Systems

27
From Brooks, “A Robust Layered Control System for a Mobile Robot”, 1985

Steels’ Mars Explorer

• Steels’ Mars explorer system, using the

subsumption architecture, achieves near-optimal
cooperative performance in simulated ‘rock
gathering on Mars’ domain:
– The objective is to explore a distant planet, and in
particular, to collect sample of a precious rock. The
location of the samples is not known in advance, but it
is known that they tend to be clustered.

28
 Steels’ Mars Explorer Rules
• For individual (non-cooperative) agents, the lowest-level
behavior, (and hence the behavior with the highest “priority”)
is obstacle avoidance:
– if detect an obstacle then change direction (1)
• Any samples carried by agents are dropped back at the
mother-ship:
– if carrying samples and at the base then drop samples (2)
• Agents carrying samples will return to the mother-ship:
– if carrying samples and not at the base then travel up gradient (3)
• Agents will collect samples they find:
– if detect a sample then pick sample up (4)
• An agent with “nothing better to do” will explore
randomly:
– if true then move randomly (5)
29

Assignment #2
• Develop a simple Subsumption
Architecture for a Ciber-Mouse Agent:
– if detect an obstacle then change direction (1)
– if at the target signal end (2)
– if not at the target then rotate to find the target (3)
– if true then moves in front/randomly (4)

30