Robotics Final Review
Algorithm
A step-by-step sequence of instructions for carrying out some task.
Agent
"anything that can be viewed as perceiving its environment through sensors and acting upon that environment through effectors."
Accessible Environment
Robot has access to all necessary information required to make an informed decision about what to do next.
Reactive control
Well suited for implementing the distributed control
Control
brain controls its actions and responds to sensory input.
Behavior Fusion
combining multiple possible candidate action/behaviors - cooperative.
Sensors
convert a physical property into an electronic signal which can be interpreted by the robot in a useful way
Proprioceptive Sensors
detect internal states, such as where your joints are.
Fixed (static) hierarchy
determined once and does not change
Pneumatic Actuators
driven by air pressure
Hydraulic Actuators
driven by fluid pressure
Dynamic hierarchy
formed based on some quality (e.g. strength)
Follower Gear
is turned by the driver (i.e., attached to wheel)
Legged Locomotion
large DOFs, challenge of stability.
Gearing Up
large gear is the driver and small gear is the follower. Gearing up gives your more speed but less torque
Passive Sensors
measures a physical property only, with a detector Ex: switches, resistive light sensors, cameras
Wheeled Locomotion
most efficient use of power, low DOFs.
Exteroceptive Sensors
obtain information from the external environment.
Active Sensors
provides own signal/stimulus, with both an emitter and a detector Ex: reflectance and break beam, ultrasound and laser
Perception
sensors and sensing allow reactive interaction with the world. They provide information about the surrounding world.
Gearing Down
small gear is the driver and large gear is the follower. Gearing down gives you more torque but less speed. If your robot is going uphill, you will need more torque.
Kinematics
study of correspondence between actuator mechanisms and resulting motion without reference to force and mass
Actuators
the actual mechanism that enables the effector to execute an action; converts software commands into physical motion
Yaw
the direction in which the body is facing, i.e., its orientation within the xy plane
Driver Gear
the gear which is being turned by a power source (i.e., attached to motor)
Deliberative control
well suited for centralized control
Differential Drive(Steering)
wheels are driven independently by separate motors -> easier control
Pitch
whether the body is tilted, i.e., its orientation within the xz plane
Roll
whether the body is upside-down or not, i.e., its orientation within the yz plane
Dynamic Environment
The world changes by itself, not only due to actions effected by the robot
Episodic Environment
The world proceeds as a series of repeated episodes
Controllable DOF
There is a direct actuator for a DOF
Kin Recognition
-Distinguishing another robot from other objects -Recognizing one's team members -Typically worth the sensory and computational cost
Feedback Control
-A means of getting a system to achieve and maintain a desired state (set point, or goal state), by continuously comparing it with the current state and minimizing the error -Desired goals have two types ->Achievement goal: no more work after final state ->Maintenance goal: ongoing active work -Goal state may be internal or external or the combination of the two -Goal could be an unreachable one, as long as the robot keeps trying
Boids Algorithm
-Algorithmic steering behaviors for animated characters. -Individual elements navigate their digital environments in a "life-like" manner with strategies for seeking, fleeing, wandering, arriving, pursuing, evading, path following, obstacle avoiding, etc. -By building a system of multiple characters, each steering according to simple locally-based rules, surprising levels of complexity emerge, the most famous example being Reynolds' "boids" model for "flocking"/"swarming" behavior. -Simple Rules: ->Separation: avoid crowding neighbors ->Alignment: steer towards average heading of neighbors ->Cohesion: steer towards average position of neighbors
Principles for BBC design
-Behaviors executed in parallel/concurrently for immediacy -Networks of behaviors used to store states -Behaviors operate on compatible time-scales -The internal, unobservable behavior and the external, observable behavior are not necessarily the same. -The observable behaviors could be the result of the interaction dynamics between the internal control and the environment. -Emergent behaviors: simple rules or behaviors interact to produce more complex outputs ->Wall following ->Flocking
Explicit communication
-Broadcast, peer-to-peer, publish-subscribe -Intentional, has cost (HW and SW) -Has to consider performance issue, what if message is lost?
Non-holonomic
-CDOF < TDOF -Most robots are non-holonomic
Redundant
-CDOF > TDOF -More than one way to move the parts
Gear Functions
-Change planes of rotation -Transfer motion -Increase/decrease speed (i.e., gearing up and gearing down) -Increase/decrease power Torque -Change direction. Idler gear only changes directions, does not affect the gear ratio (speed)
Effectors
-Comprises all the mechanisms through which a robot can effect changes on itself or its environment -Categories: manipulator and mobile
Distributed Control
-Control is spread over multiple/all members of the team -Each robot uses its own controller to decide what to do -No central information gathering, no bottlenecks -Works well with large teams, doesn't slow down with size
Tightly Coupled Coordination Strategy
-Cooperate on a precise task using communication, turn-taking. -Dependent on each other, with improved group performance -Less redundancy and less robustness e.g. soccer playing, moving in formation, transporting objects, etc
Heterogeneous Teams
-Different, non-interchangeable members -Typically requires active cooperation in order to produce coordinated behavior
Behavior Based Control
-Distribute thinking over acting, think the way you act -Involving the use of "behaviors" as modules for control. Incorporating the best of reactive systems, but does not involve a hybrid solution -What is a behavior? ->Achieve, maintain particular goals (e.g. homing, wall-following). ->Time-extended, not instantaneous ->Take inputs from sensors, and other behaviors. Send outputs to effectors and other behaviors. ->More complex than actions (e.g. find-object vs. stop-turn-left)
Reactive Control
-Don't think, don't look ahead, just react! -Operates on a short time scale -Based on a tight loop connecting the robot's sensors with its effectors -Purely reactive controllers do not use any internal representation; they merely react to the current sensory information -Collection of rules that map situations to actions ->Simplest form: divide the perceptual world into a set of mutually exclusive situations, recognize which situation we are in and react to it ->(but this is hard to do!)
Levels of Processing
-Electronics (low level): such as measuring voltages -Signal processing (medium level): such as separating voice from noise -Computation (high level): such as recognizing an object from an image
RoboCup
-First RoboCup was held in Nagoya, Japan, during IJCAI-97. -2013 was held in the Netherlands, 2014 in Brazil
Hybrid control
-Good for both the centralized and distributed control -The centralized controller performs the SPA (sense-plan-act) loop, individual robots monitor their sensors and update the planner.
Behavior-based control (BBC)
-Good for implementing the distributed control -Each robot behaves according to its own local BBC controller
Homogeneous Teams
-Identical (in form and/or function), interchangeable members -Could be coordinated with simple mechanisms, may require no intentional cooperation to achieve effective group behavior (such as emergent flocking)
The need for communication in a team
-Improving perception -Synchronizing action -Enabling coordination and negotiation
Implicit communication
-Individual robot leaving information in the environment -Stigmergy - information is conveyed through changing the environment, such as ant trails (pheromone left by ants). -Positive feedback: amplifying effects, in contrast to the regulatory feature of the negative feedback control
Challenges of Teamwork
-Interference and collisions between robots -Communication collisions -More uncertainty with more robots roaming around -More expensive
Deliberative Control
-Look-ahead; think, plan, then act -Planner-based architecture ->Sensor (S) ->planning (P) ->Acting (A) -Requirements ->Lots of time to think ->Lots of memory ->(but the environment changes while the controller thinks)
Switches
-Measuring current to detect an open or closed circuit -Used as a contact sensor, limit sensor, or shaft encoder sensor. -Could be implemented as a surface contact or a whisker (antenna) like structure
Examples of what could be communicated in foraging
-Nothing (could still work well in merely coexisting) -Task-related state: locations of objects, # of recently seen robots, etc -Individual state: ID #, energy level, # of objects collected, etc -Environment state: blocked paths, dangerous conditions, new-found shortcuts, etc -Goal(s): direction to the nearest object, etc Intentions: I'm going that way because ...
Degrees of Freedom
-Number of directions in which a robot can be controlled -Three for position (x, y, z - Translational DOF) -Three for orientation (roll, pitch, yaw - Rotational )
Roomba Motors
-One driving each wheel (2 total) -One driving the vacuum -One driving the spinning side brush -One driving the agitator assembly
Light Sensors
-Photocells convert light intensity to resistance in the circuit -Work even with invisible light (such as infrared) -Could be used for measuring intensity, differential intensity or break in continuity -Reflectance sensors: active sensors with emitter and detector side by side -Break beam sensors: emitter and detector face each other -Calibration is used to reduce noise
Types of Feedback Control
-Proportional Control (P): gain is proportional to error, and it determines the amount of oscillating or damping -Proportional derivative control (PD) -Proportional integral derivative control (PID)
RCX
-Robotics Command Explorer -Hitachi h8300 microprocessor
Roomba's Decision Making
-Roomba uses iRobot's AWARE Robotic Intelligence System to make many decisions for itself, so minimal human input is required. -The AWARE system is made up of multiple sensors that pick up environmental data, send it to the robot's microprocessor and change Roomba's actions accordingly. -Roomba's sensors allow it navigate your home with relative autonomy.
Behavior Arbitatraion
-Select one action/behavior from multiple possible candidates - competitive (one choice wins). ->Fixed priority hierarchy (pre-assigned priorities) ->Dynamic hierarchy (behavior priorities change at run-time)
Issues with Sensors
-Sensors produce signals, not symbols. -Signal-to-symbol problem: from sensor input to intelligent response which requires an abstract and symbolic form
Centralized Control
-Single, centralized controller takes information from all other robots, thinks, sends commands to all -Is slow and gets slower when the team size increases -Not robust and the centralized controller is a bottleneck of the whole system -Advantage: optimal solution to a given problem
Why Teams?
-Some jobs require more than one robot -Better, faster, cheaper -Can be in multiple places at once -Redundancy
Algorithm Features
-Speed (number of steps) -Memory (size of work space) -Complexity (can others understand it?) -Parallelism (can you do more than one step at once?)
Roomba Sensors
-The cliff sensors constantly send out infrared signals, and Roomba expects them to immediately bounce back. If it's approaching a cliff (steps), the signals get lost and the Roomba knows to head the other way. -hen Roomba knocks into something, its bumper retracts, activating mechanical object sensors that tell Roomba it has encountered an obstacle. It then backs up, rotates, and moves forward until it finds a clear path. -The wall sensor is located on the right side of the bumper that detects the bounced-off infrared signal sent out from an emitter on the left side. It lets Roomba follow very closely along walls and around objects (like furniture) without touching them. -It has dirt sensors in order to figure out which areas need more cleaning. When the agitator kicks up a large amount of dirt, it causes more vibration when it hits the metal plates of the sensors. The sensors detect that acoustic increase and tell Roomba to go over the area again.
Holonomic
-The number of controllable DOF (CDOF) is equal to total number of DOF (TDOF). -i.e. CDOF = TDOF
Gait
-The way a robot moves by using a particular pattern of footfall -2 legged: alternating swing and stance phases. -4 legged: lateral walking vs. diagonal walking -6 legged: alternating tripod gait vs. ripple gait.
Robot History
-The word robot came from the Czech word robota, which means slave. -Used first by playwright Karel Capek, "Rossum's Universal Robots" (1923) -The Three Laws of Robotics ->A robot may not injure a human being, or, through inaction, allow a human being to come to harm. ->A robot must obey the orders given to it by human beings except where such orders would conflict with the First Law. ->A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Hybrid Control
-Think but still act quickly, both work in parallel -Use the best of both worlds (deliberative and reactive) -Combine open-loop and closed-loop execution -Combined different time scales and representations -Typically consists of three layers: ->Reactive layer ->Planner (deliberative layer) ->Integration layer to combine them ->(but this is hard to do!)
RoboCup Soccer
-Ultimate goal: a fully autonomous humanoid robotic soccer team to beat human World Cup Champions by the year 2050. -Leagues: --Standard Platform league (Sony's Aibo -> Aldebaran Robotics' Nao) --Small size league (5 robots of <18cm diameter and <15cm height) --Middle size league (5 robots, each fits a 50x50x80cm3 box) --Simulation league (software) --Humanoid League
Wheels and Steering
-Wheels are the choice of locomotion in robotics because they are highly efficient and simple to control -Most wheeled robots are not holonomic ->Following a specific trajectory (motion planning) is more difficult than simply moving from one place to another (navigation).
Torque
-a measure of how much a force acting on an object causing that object to rotate, or, "Twisting Force" -Torque must overcome friction to move the wheels of a vehicle (gravity too if it is on an incline).
Multitasking
-a method by which multiple tasks, also known as processes, share common processing resources such as a CPU. -In the case of a computer with a single CPU, only one task is said to be running at any point in time, meaning that the CPU is actively executing instructions for that task. -Multitasking involves scheduling which task may be the one running at any given time, and when another waiting task gets a turn.
Electric Motor Actuators
-affordable, small, and simple -DC Motors: speed of motor controlled by input voltage -Servo Motors: position controlled (-180° ~ 180°)
Robot
-autonomous embodied agent -Contains sensors to perceive its own state and its surrounding environment -Possesses effectors which perform actions -Has a controller which takes input from the sensors, makes intelligent decisions about actions to take, and effects those actions by sending commands to motors
Loosely Coupled Coordination Strategy
-group recognition, simple coordination, -don't depend on each other, robust, -difficult to do precise tasks -Well-suited for foraging, herding, distributed mapping, etc
Disadvantage of Distributed Control
-hard to design individual behavior so that they will work well in their interactions to produce the designed group behavior (see competitive soccer playing). -Statistics tools can be used when there are many components and they are simple. -In robotics, we have small number of complicated components. Thus we have to solve the "inverse problem" - going from the global behavior to the local rules.
Idler Gear
-is placed directly between two others (i.e., between driver and follower). -It does not affect the gear ratio (rotation speed). Idler gear only changes direction.
Merely Coexisting Coordination Strategy
-no communication or even recognition of each other (seen as obstacles). -Interference increases with the # of members. -Well-suited for foraging, construction, etc
Control architecture
-provides a set of principles for organizing a robot's control system -Provides constraints, guides how programs are structured -Refers to software control level, not hardware! -Implemented in a programming language
Static Stability
-robots maintain upright without constant active control -Maintained when the center of gravity (COG) is above the polygon of support -Statically stable walking is slow, energy inefficient
Dynamic Stability
-robots must actively balance or move to maintain stability -The inverse pendulum model for one legged balance -Two legged walking alternates between swing and stance phase between the two legs.
Power
-the amount of work that can be done in a certain amount of time Translational: Power = Force X Speed -Rotational: Power = Torque X (2π X Rotational_Speed) -The maximum power of a motor is fixed -Tradeoff between Torque and Speed
Representation
-the form in which information is stored or encoded in the robot -Representation of the world is called a world model. E.g. a map in navigation, many ways of modeling exist: ->Odometric path: go fwd for 3cm, turn left 90 deg... ->Landmark-based path: turn left at the first junction... ->Topological map (landmark-based map, the order of the junctions is not required) ->Metric map: drawing the map with exact distances. -The world model needs to be kept accurate and updated. Some are long-lasting, others are transient -Representation comes with a cost (sensing, computation, memory), and the level of its use depends on the architecture chosen.
How to Solve an Algorithm
1-Understand the problem 2-Devise a plan 3-Carry out your plan 4-Examine the solution
Deterministic Environment
Any action that a robot undertakes has only one possible outcome.
Types of Effectors
Arm, leg, wheel, gripper, etc
Sensor Fusion
Combining multiple sensors to get better information about the world.
Discrete Environment
Sensor readings and actions have a discrete set of values
Gear Ratios
You can calculate the "gear ratio" by using the number of teeth of the "driven gear" divided by the number of teeth of the "driver gear".