Out line Robot ics Percept ion Robot ics Planning Reading: - PDF document

Out line • Robot ics – Percept ion Robot ics – Planning • Reading: R&N Sect . 25.1-25.4 J uly 26, 2005 CS 486/ 686 Univer sit y of Wat erloo 2 CS486/686 Lecture Slides (c) 2005 P. Poupart Robot s Robot s • Mobile r obot s – Shakey t he robot (SRI 1968) • Manipulat ors • First mobile robot – Service robot s (CMU’s Minerva) – Physically anchored • Museum t our guide robot – Most indust rial robot s – Unmanned land vehicle (NavLab) • Aut onomous highway driving • Assembly lines – Unmanned air vehicles • Surveillance, crop-spraying, milit ary operat ions – Aut onomous underwat er vehicles • Deep see explorat ion – Planet ary rovers (Nasa’s Soj ourner) 3 4 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart Shakey t he robot Nasa’s Soj ourner • First mobile robot • Planet ary (SRI 1968) rover 5 6 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart 1

Honda’s P3 and Asimo Robot Hardware • Humanoid • Sensor s robot s – Range f inders (sonars, lasers) – Tact ile sensors – GPS – I maging (video cameras) – Propr iocept ive sensors (odomet ry) – Microphones 7 8 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart Range scan example Robot ic Percept ion • Challenge: noisy sensor s • What st at ist ical model should we use t o inf er t he st at e of t he world? • HMMs (or DBNs) s 0 s 1 s 2 s 3 s 4 o 1 o 2 o 3 o 4 9 10 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart Simult aneous mapping and localisat ion Robot Localisat ion • ht t p:/ / robot s.st anf ord.edu/ movies/ mapping1-new.avi • Sebast ian Thrun • Sebast ian Thrun • ht t p:/ / robot s.st anf ord.edu/ movies/ sca80a0.avi 11 12 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart 2

Robot Hardware Degrees of Freedom • Ef f ect or s • A robot has one degree of f reedom f or each independent direct ion of movement – Revolut e j oint s and prismat ic j oint s – Grippers P – Wheels R R – Legs R R – Speakers R 6 degrees of freedom 13 14 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart Degrees of Freedom Degrees of Freedom • Holonomic r obot s: • How many degr ees of f reedom (DOF) – # ef f ect ive DOF = # cont rollable DOF does a car have? – Most robot arms – 3 ef f ect ive DOF: x, y, orient at ion – Easy t o cont rol – 2 cont r ollable DOF – Complex mechanics • Non-holonomic robot s – # ef f ect ive DOF > # cont rollable DOF θ – Most mobile robot s – Harder t o cont rol (x,y) – Simple mechanics 15 16 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart Planning POMDP (or DDN) • Challenge: • Graphical Represent at ion – Noisy sensors – Uncert ain act ion ef f ect s • What st at ist ical model can we use? o 2 o 3 o o o 1 – Part ially observable Markov decision a 3 a 2 a 0 a 1 process (POMDP) s 0 s 1 s 2 s 3 s 4 – Dynamic decision net work (DDN) r 2 r 3 r 4 r 1 17 18 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart 3

Greedy Pat ient Finding POMDP P at ient Finding • Sebast ian Thrun • Sebast ian Thrun • ht t p:/ / robot s.st anf ord.edu/ movies/ good.people.avi • ht t p:/ / robot s.st anf ord.edu/ movies/ bad.people.avi 19 20 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart Next Class Ot her subf ields of Robot ics • Next Class: •Cour se wrap up • Mechanics •Final exam inf o – Hardware engineering •Ot her AI courses • Cont rol – Form of planning – Mainly concerned wit h “st abilit y” 21 22 CS486/686 Lecture Slides (c) 2005 P. Poupart CS486/686 Lecture Slides (c) 2005 P. Poupart 4