Robotics and Embodied Cognition Overview

Reminders

• Sign in to AttendanceRadar for a quiz.
• Short paper #2 is due Tuesday, November 19th.

Robotics and Embodied Cognition

• Cognition changes when situated in a physical environment.
• Cognitive agents need to be designed for physical environments.
• Intelligence is the things that highly educated scientists find challenging (e.g., chess, complex math).
• Simple tasks (e.g., walking, object recognition) were not considered intelligent until recently.

Embodied Cognition (Page 8)

• Traditional approach: the mind is separate from the environment, environment provides inputs and outputs of the mind.
• Embodied cognition: the mind and environment together are a single cognitive system.

Embodied Cognition (Page 9)

• Refers to a spectrum of ideas, including:
  • "Simple" embodied cognition: cognitive systems operate in the physical world.
  • "Medium" embodied cognition: the world and body shape thoughts.
  • "Radical" embodied cognition: the mind can't be meaningfully studied in isolation from the world.

Modal Representations (Page 10)

• A branch of embodied cognition: most/all semantic representations are modal (tied to sensory and motor information associated with the concept).
• To study this, researchers look for embodied information even when not needed for the task.

Reaction Time to Objects (Page 11)

• Researchers measured reaction times for different locations on objects (ex. handles on cups, etc).
• Reaction time data from study showed that reaction times were faster when relevant to the object.

Where is Yesterday? (Page 12)

• A concept related to remembering things.

Time Perception (Page 13 and 14)

• Some studies have shown that people's perception of time depends on the context (with and without a visual object).
• Studies showed differences in time perception depending on language (English or Hebrew) and the time interval.

Why the Environment is Part of a Cognitive System (Page 15)

• Need to understand why we need to consider the environment as part of a cognitive system.

Programming a Robot (Page 16)

• Move forward whenever possible
• If stuck, back up a bit and turn a random amount
• Task: navigating, overcoming obstacles

Programming a Robot (Page 18)

• In a specific environment, the robot's function is to pile up blocks.

Robot Obstacle Avoidance (Page 19)

• If obstacle is in left rear or right front, turn left.
• If obstacle is in right rear or left front, turn right.
• Task is to navigate a maze.

Marr's Highest Level (Page 22)

• To study a system's functional level, the robot and its environment are relevant for analysis.
• "Radical" embodied cognition argues cognitive agents aren't studied in isolation.

Off-loading Cognition (Page 23)

• Robots can get away with limited onboard cognition.
• Humans strategically use the environment to improve cognition.

Australian Aboriginal Memory Technique (Page 24)

• Use of physical locations to associate information.
• Walking through these locations enables remembering.
• The method uses landmarks to remember things.

Memories and Places (Page 25)

• Memories can be triggered by returning to places reminiscent of past experiences.

Dynamical Coupling (Page 26)

• Embodied cognition is an ongoing interplay between robots and their environment.
• This contrasts with representation-hungry (more complex) systems.
• Systems can repeatedly perform simple calculations through observing the environment.

Catching a Baseball (Page 27)

• Strategies for catching include measuring position, velocity, spin, and wind speed.
• Physics simulations are used to predict the ball's landing point.
• Adjusting running direction to keep the ball moving in the same direction.

Rodney Brooks (Page 29)

• Actionist/situated/behavioral robotics: against maintaining representational states inside robots.
• Focus is on producing behavior through interactions with the environment.
• The world is its own best model (environment is complete in every detail).

Robot Control Layers (Page 30)

• Finite state machines are layered for robot control.
• Each layer builds atop the ones below.
• Layers aren't dependent on each other.

Social Cognition (Page 31)

• Other people/robots are considered part of the environment.
• Cognition can involve groups of minds.

Robot Computing (Page 32)

• A robot increases a counter by 1 every 30 seconds.
• When the counter reaches 100, it flashes a light and resets the counter to 0.
• Seeing a nearby flash adds 10% to the counter.
• Determining if actions are calculated by a robot.

Consensus Synchronization (Page 34)

• An algorithm can enable group consensus synchronization.
• The cognitive system is the whole group, not the individual.

Robot Navigation (Page 35)

• Robots turn to align their direction and position with the average of nearby robots.
• The function is related to animal flocking or bird flight.

Autonomous Driving (Page 36)

• Autonomous driving models decision-making at a multi-car level.
• Different types of interactions can be planned by vehicles, and the result can impact decision-making.

Robot Morality (Page 37)

• Robots can take actions in the physical world and cause harm to people, animals or property.
• There are cases where fast decisions can need to be taken without human intervention.

Asimov's Laws of Robotics (Page 38)

• First Law: Robots must not injure humans.
• Second Law: Robots must obey human orders, except if that conflicts with the First Law.
• Third Law: Robots must protect themselves as long as it does not conflict with the first two laws.

Self-Driving Car Dilemma (Page 40/41)

• Dilemmas arise when self-driving cars face situations requiring a choice between harming one or many people.

Summary (Page 42)

• Cognitive systems in the physical world face challenges in simple actions.
• Abstract representations are influenced by physical interactions with the environment.
• Environments can be tightly coupled with systems, algorithms, and representations.

Explore the concepts of embodied cognition and its implications in robotics. This quiz covers the fundamental ideas around how cognitive agents interact with their physical environments and the evolution of what constitutes intelligence. Delve into the spectrum of embodied cognition from simple to radical approaches.