Autonomous systems notes

Robot

A robot is an autonomous system that exists in the physical world, can sense it’s environment, and can act on it to achieve some goals

Robotics

The study of robots and their autonomous and purposeful sensing and acting in the physical world.

Teleoperated

Machines that are externally controlled by humans and are not true robots.

Autonomous

An autonomous robot acts on its own decisions and is not controlled by humans.

Control theory

The mathematical study of automated control systems.

Cybernetics

The study and comparison of communication and control processes in biological and artificial intelligences.

Biomimetic

Machines with properties similar to those of biological systems.

Reactive control

A means of controlling a robot using a collection of prioritized reflexes.

Artificial alife

A field related to robotics that focuses on computational systems within a computer.

Photophilic

Attracted to light.

Photophobia

Repulsed by light.

Excitatory connection

The stronger the input, the stronger the output.

Inhibitory connection

The stronger the input, the weaker the output.

Vision-based robot navigation

Using vision to move a robot around.

Physical embodiment

Having a body in the physical world. Having a physical body means having to be aware of other bodies and objects around you

Sensors

Physical devices that allow a robot to perceive its environment in order to get information about itself and its surrounding

Sensing/perception

The process of receiving information about the world through sensors.

Robot's state

The description of a robot at any point in time.

State space

All possible states a system can be in

External state

State of the world as the robot can perceive it

Internal state

State of the robot as the robot can perceive it

Representation/internal model

Remembering information about the world.

Effectors

Enable a robot to take action or do physical things.

Actuators

Muscles or motors used to do actual work.

Locomotion

Moving around/going places

Manipilation

Handing objects

Degrees of freedom (DOF)

The dimensions in which a manipulator can move.

Controllers

Provide the hardware/software that make the robot autonomous by using the sensor inputs to decide what to do

Autonomy

The ability to make your own decisions and act on them.

Things robot can learn

1) A robot can learn about itself (how to control its actuators, how to

process data from the sensors.

2) A robot can learn about its environment. (mapping the area)

3) Robot can learn about other robots. (working collaboratively)

Reinforcement learning

Learning by trying different things and seeing what happens. Getting punished or rewarded for the specifc action

Exploration

The process of trying all possible actions in all states.

Exploitation

Using what has been learned.

Control policy

A complete state-action table.

Value function

The value of being in each state relative to the goal.

Positive feedback

Reward.

Negative feedback

Punishment.

Unsupervised learning

Learning without an external supervisor.

Neural network learning

Learning algorithms using neural networks.

Statistical neural networks

Learning with formal techniques from statistics and probability.

Lifelong learning

Continually learning and constantly improving.

Clustering

Grouping similar data together.

Parameters

Trainable weights of a neural network.

Hyperparameters

Settings that describe the neural network and learning settings.

Autoencoding

Reducing the dimensionality of data by compressing it.

Local optimum

A solution that is not the global optimum but cannot be improved using specific change parameters.

Simulated annealing

Accepting improving solutions but also selecting new ones with lower fitness values.

Evolutionary algorithms

Algorithms based on evolution theory.

Selfish gene hypothesis

Genes using mortal individuals as hosts to propagate themselves.

Optimization problem

Finding the best solution in a search space.

Heuristics search

Starting with random solutions and evaluating them.

Genotype

Set of genes representing a chromosome.

Phenotype

The actual physical representation of the chromosome.

Mutation

Changing an individual to explore the representation space.

Bit string mutation

Changing a bit to its opposite.

Mutation for real numbers

Applying the same techniques as with bit string mutation.

Mutating for ordered representation

The process of changing the values in an individual while respecting the constraints of the problem. Swap of 2 values on 2 different loci should be used

Recombination

The process of combining genetic material from multiple parents to create offspring.

L-point crossover

A type of recombination operator where a single cutting point is generated and both individuals are cut at that point to create two possible children.

Uniform crossover

A type of recombination operator where the value for each location in the offspring is randomly chosen from either the first or second individual.

Fitness-proportional selection

Parents that are allowed to reproduce themselves are assigned a probability for reproduction that is based on their fitness

Selection strategy

Determines how individuals of a population are chosen for generating offspring

Tournament selection

A selection strategy where a group of individuals is randomly selected and the best one from the group is used to create an offspring.

Fitness proportional selection

A selection strategy where the probability of an individual reproducing is based on its fitness value.

Schemata-theorem

If the crossover operator doesn't destroy good building blocks too often they can be quickly mixed and stay in the population with less diverse population the effect of crossover diminishes

Rank-based selection

A selection strategy where individuals are assigned ranks and the probability of an individual becoming a parent is based on its rank.

Elitism

A technique in genetic algorithms where the best individuals from one generation are carried over to the next generation to ensure that the quality does not decrease.

Cellular automata

Decentralized spatial systems with simple components that are locally connected and update synchronously.

ROS (Robot Operating System)

A distributed computing environment that allows for communication between nodes in a robotic system.

Node

A single-purpose program that either "publishes" on "subscribes" to information on a certain topic

Passive actuation

Utilizing the potential energy in the mechanics of an effector and its interaction with the environment instead of active power consumption.

Electric motors

Motors that use electric current to generate motion.

Hydraulics

Actuation system that uses fluid pressure to generate motion.

Pneumatics

Actuation system that uses air pressure to generate motion.

Photo-reactive materials

Materials that perform physical work in response to the amount of light around them.

Chemically reactive

Materials that react to certain chemicals and perform physical work.

Linear actuator

An actuator that provides linear movement, getting longer or shorter.

Thermally reactive

Materials that change in response to temperature.

Piezoelectric

Materials that generate electric charges when pushed or pressed.

Backlash

Any looseness between meshing gears that causes the gear to move sloppily back and forth between the teeth.

Servo motors

Motors that can turn their shaft to a specific position.

Pulse-width modulation

A method of controlling the amount a motor shaft turns based on the duration of a pulse.

Position control

Controlling the motor to track a desired position at all times.

Torque control

Controlling the motor to track the torque at all times.

Degree of freedom

The minimum number of coordinates required to completely specify the motion of a mechanical system.

Transitional DOF

The degree of freedom that allows the body to translate without turning. (X,Y,Z)

Rotational DOF

The degree of freedom that allows the body to rotate. (Roll, pitch, yaw)

Roll

Rolling from side to side

Pitch

Pitching up and down

Yaw

Yawing (turning) left and right

Holonomic

When the number of controlled DOF is equal to the total DOF.

Non-holonomic

When the number of controlled DOF is less than the total DOF.

Redundant

When the number of controlled DOF is more than the total DOF.

Static stability

The ability of a robot to stand still without falling.

Polygon of support

The area covered by the ground points, which determines the stability of a robot.

Statically stable walking

The ability of a robot to walk while staying balanced.

Dynamic stability

The body of the robot actively balances or moves to remain stable.

Inverse pendulum

Balancing a one-leg robot.

Gait

The particular way in which a robot moves, including the order in which it lifts and lowers its legs and places its feet on the ground.

Desirable gaits

Stability, speed, energy efficiency, robustness, and simplicity.