CMPUT 412 Introduction Csaba Szepesvári University of Alberta
Table of contents Admin Robots Basics of control Robot design
Admin o Teams: n n o o Not yet assembled (next week) Rotational scheme First lab: Learn about NXT Evaluation update: n n Team members evaluate each others performance Fill out a form (who did what) Team of N can get total of N*5 points (5: good), consensus 10% - class participation (M=0. 9*T + 0. 1*cp)
Robots Manipulators Mobile robots
Manipulators
Here: Mobile Robots o Mobile: o Changes its point of reference with respect to the environment Consequences: n n Advantages Problems: o Where am I? Hidden state, aliasing, . . o What is happening? Unstructured environments. .
Mobile robots: Advantages o o o Provide specialized access n hazard, environment (no air, etc. ), distance/time (Mars) Reduce operating costs n lower overheads, reduced maintenance costs (gentler treatment of the machinery) Increase productivity n “permanent” availability - more hours; higher throughput Improve product quality n accuracy, consistency New human services! n human interactivity
Applications o o Medical Cleaning n n o o o Commercial Household Sales (automated gas pump in Japan) Agriculture (Demeter) Forestry (pruning Xmas trees) Lawn care (golf courses) o o o Hazardous (high-power line inspection, pipe inspection, building inspection) Mining/Excavation Space / Undersea Military Security Personal (handicapped aids; @ home) Entertainment (Sony’s Robo. Dog)
AGV by Volvo o Automatic Guided Vehicle of VOLVO Goal: transport motor blocks from on assembly station to an other Navigation: n n o guided by an electrical wire installed in the floor able to leave the wire to avoid obstacles. Over 4000 AGV only at VOLVO’s plants!
Helpmate o o Goal: transportation in hospitals Navigation: Look at the ceiling with a camera (landmarks) + corners
BR 700 Cleaning robot o o o Goal: Cleaning Navigation: sonar system + gyro Kärcher Inc.
ROV Tiburon underwater robot o o o Goal: Underwater archeology Control: Hovering only Navigation: Teleoperated
The Pioneer o o Goal: Explore sarcophagus at Chernobyl Control: Teleoperated
Pioneer I. o o Goal: Research Stanford Robotics Institute
The B 21 Robot o o Goal: Research Manufacturer: Real World Interface
Forester robot o o o Moving wood out of forest Control: Human, leg coordination automated Pulstech
Tube inspection o o o EFPL robot (4 mins) Goal: sewage tube inspection and reparation Control: teleop. HÄCHER Goal: Airduct inspection Control: Teleop. EFPL
Sojourner – Mars robot o o Goal: Explore Mars Control: Teleoperation (1 cm/sec, 10 min delay), some obstacle avoidance (2 mins)
Nomad • Goals: autonomous search of Antarctic meteorites • Developer: CMU, NASA • 1997
Honda
Aibo from Sony Stereo mics Color camera 10 cm
Vacuum Cleaners Dyson DC 06 (cancelled 2004) i. Robot Roomba Electrolux Trilobite
The Cye Personal Robot o o o Goal: Vacuum cleaner, trailer Remote controlled (PC) Two wheels!
How to control robots? o. Controlled systems o. Controller goals
Abstract control model Environment Sensations (and reward) actions Controller = agent “Perception-action loop”
Zooming in. . memory reward external sensations agent state internal sensations actions
Mathematical model o o Plant (controlled object): n xt+1 = f(xt, at) + vt xt : state n Yt = g(xt) + wt yt : sens/obs State: Sufficient statistics for the future n Independently of what we measure. . or. . n o Relative to measurements (goals) Controller n at = F(y 1, y 2, …, yt) at: action/control => PERCEPTION-ACTION LOOP “CLOSED-LOOP CONTROL” o Design problem: F = ?
What: Goals of a robot o Goal directed behavior n n n o o Environmental (manipulation) goals o Brew coffee o Pour water into the coffee machine o Pick up a cup Position goals: o Get to a place o Deliver packages Homeostasis Reward hypothesis: Homeostasis: t=1 T rt ! max zt 2 Z, zt: internal signals
Competencies: What -> How o High level goals n n n o environmental (manipulation) goals homeostasis position goals Supporting competencies n knowledge of o o o environment internal state position
How: Controllers o Feedforward: n n o a 1, a 2, … is designed ahead in time ? ? ? Feedback: n n n Purely reactive systems: at = F(yt) Why is this bad? Feedback with memory: mt = M(mt-1, yt, at-1) ~interpreting sensations at = F(mt) decision making: deliberative vs. reactive
Feedback controllers o Plant: n n o Controller: n n o o xt+1 = f(xt, at) + vt yt+1 = g(xt) + wt mt = M(mt-1, yt, at-1) at = F(mt) mt ¼ xt: state estimation, “filtering” difficulties: noise, unmodelled parts How do we compute at? n n With a model (f’): model-based control o. . assumes (some kind of) state estimation Without a model: model-free control
A control scheme for mobile robots Knowledge, Data Base Mission Commands Localization Map Building "Position" Global Map Cognition Path Planning Path Execution Raw data Actuator Commands Sensing Acting Motion Control Path Information Extraction Perception Environment Model Local Map Real World Environment (Siegwart, Nourbakhsh)
How to Design Robots? Mechanics - Body (chassis, actuation) Energy - Power Inputs - Sensors Brain - Control
Deconstruction. . o Mechanical n n o chassis – load bearing & distribution, suspension effectors (mobility) wheels/gears, legs, tuna!, helicopter rotors pan/tilt units; reconfiguration units (pipe crawlers, throw-in-bldg) Power n n Batteries Solar Nuclear Generators (gas)
Deconstruction. . /2 o Sensing n n internal sensors - voltages, load, encoders position estimation sensors – measuring relative position of robot o n environmental sensors - measuring characteristics of the external world o n o compass, odometer, gyro, accelerometer, INS, GPS, startscope, sunscope bumpers, sonar, infrared, laser-rangefinder, radar, CCD’s, Mars soil sampler weird ones: torque/force transducers, limit switches Control n n in most general case: y* a multilevel control versus 1 -level control: “goal” motion; then achieving it. . (outdoor mobile robot)
Big issue questions o o o Scaling up & scaling down! Sensors: how much is enough Exceptions/errors n o Completeness/optimality n o What are they? What does it mean? ? ? What do they mean? Deliberation vs. reactivity n Models/lookahead required? Memory?
Big issue questions/2 o Representation n o How do we do uncertainty? n o o Model explicitly, engineer around? Better sensors? Better effectors? Learning (how does it fit in? ) n o If we want one? transparency/opacity Off-line learning / parameter tweaking, RL? Robot teams ? ? Architecture! n tiered, parallel, sequential, etc.
Summary o Mobile robots n o Working with (mobile) robots requires understanding of: n n n o Advantages and challenges Engineering issues Control issues Signal processing issues Closed-loop, FB vs. FF control, memory, state-estimation/filtering, modelbased/-free, deliberative vs. reactive
Скачать презентацию CMPUT 412 Introduction Csaba Szepesvári University of Alberta
1931e8df51d027582c74642a00acd0e5.ppt