Robotic Architectures and Intelligence Quiz

Architectures for Autonomy

• Autonomy has a different programming style than traditional programming
• Questions arise about how much AI is needed for a robot
• Can intelligence be added in layers, like upgrading to a "pro" version or downloading "apps"?

Specific Learning Objectives

• Apply Levis' definition of architectures to organizing software in an intelligent robot.
• Name and describe the layers (behavioral, deliberative, interface) within a canonical operational architecture of an intelligent robot.
• Name the four primitives of robotic intelligence (sense, plan, act, learn).
• List vulnerabilities where poor choices or assumptions can lead to failures in the operational architecture.

Types of Architectures

• Operational architecture: describes what a system does, not how it does it.
• Systems architecture: describes how a system works in terms of major subsystems.
• Technical architecture: describes how a system works in terms of implementation details and language.

Organizing Software

• An overall style of design or organization is called an architecture.
• It provides a principled way of organizing a control system, but it may impose constraints on how a control problem can be solved.
• An architecture describes a set of architectural components and how they interact.

Focus of this Course

• The course focuses on operational architecture, which describes what a system does not how it does it.
• Systems architecture describes how a system works through its major subsystems.
• Technical architecture describes the system using implementation details and language.

OODA Loop

• Observe, Orient, Decide, and Act; used by military pilots
• Not everything intelligent requires deciding, some behaviors are reflexive or rely on muscle memory.

Action-Perception Cycle

• Action-oriented perception is similar to OODA but has a stronger scientific basis.
• AI has cast behaviors as primitives: SENSE, PLAN, ACT and LEARN.

Canonical Architecture

• "Upper brain" (cortex) - reasoning over symbols (information), regarding goals
• "Middle brain" - converting sensor data into symbols (information)
• Spinal cord and "lower brain" - skills and responses

Canonical Al Robotics Operational Architecture-Attributes for Describing Layers

• The description and justification for software functions in each layer are based on five attributes: primitives, perceptual ability, planning horizon, time scale, and model use.

Primitives for Robot Intelligence

• Sense
• Plan
• Act
• Learn

Robot Primitives (Inputs and Outputs)

• Sense: Input is sensor data, output is sensed information.
• Plan: Input is information, output is directives.
• Act: Input is sensed information or directives, output is actuator commands.

Why Worry About Software Organization?

• Artificial Intelligence has many distinct areas with their own algorithms and data structures that need to be integrated.
• Software engineering is crucial for a successful software enterprise.

Thinking About Architectures is Good Software Engineering

• Abstraction: ignores details to focus on general organization of intelligence.
• Modularity: high cohesion (do one thing well), low coupling (which means may be able to add "apps").
• Anticipation of change, incrementality: adapt to support evolution.
• Generality: not reinventing the wheel each time.

What Does This Mean for Programming?

• Transition from reactive to deliberative programming.
• Impacts computer vision (direct and recognition-based) involves two kinds of perception .
• Different time horizons (present, past, future).
• A central system (a world model) is necessary for combining symbols, history, knowledge.
• Theory of mind—interaction involves beliefs, desires, intentions (BDI) and common ground.

Can Intelligence be Added in Layers?

• Intelligence is organized in layers.
• Libraries of algorithms ("logical ") can be created.
• Adding behaviors in layers is often complex and non-trivial, especially coordination between layers.
• Capabilities are not always easy to add incrementally.

How Much Artificial Intelligence Does a Robot Need?

• Depends on the robot's functions (generate, monitor, select, implement, execute behaviors), learn, planning horizon, and update speed.
• Requires consideration of local or global models.

Other Ramifications: Planning Horizon, Time Scales

• Different time scales are important for programming decisions (present, past, present+past, present+past+future), very fast and slow

Canonical Operational Architecture

• Each layer has a different programming style in organization
• Layer organization does not typically consider interaction details.

Programming Languages

• Procedural, functional, ontological languages (such as OWL) are used with interaction layer.
• Functional languages such as Lisp are used with the deliberative layer
• Procedural languages such as C, C++, Java are used with the behavioral layer

Advantages of Programming in Layers with Different Styles

• Decomposition of a complex system.
• Matching right tools and mindsets for the task.
• Add to working, verified code.

3T Hybrid Operational Architecture

• Three-tier operational architecture
• Five Attributes
• Primitive (sense, plan, act)
• Perception (direct, recognition)
• Planning horizon (present, past-future)
• Time scale (very fast, fast, slow)
• Models (local, global)

Future Questions and Lectures

• Examine levels of autonomy
• Understand systems architecture of hybrid models
• How do 7 areas of AI fit in?

Other Observations

• Current practice is good with deliberative on symbols and behaviors.
• A key barrier is converting sensory data to symbols.
• Understanding human intent is another major barrier.

Summary and Additional Thoughts

Summary: Architectures

Other Observations

Test your knowledge on the functions and structures of robotic architectures. This quiz covers topics such as operational architecture, software engineering in robotics, and the layers of robot intelligence. Dive into the complexities of robotics and assess your understanding of key concepts.