Algorithms for Sensor-Based Robotics: Introduction and Background

Transcription

1 Algorithms for Sensor-Based Robotics: Introduction and Background Computer Science 436/636 Greg Hager Simon Leonard With slides borrowed from Zach Dodds and Howie Choset

2 What is a Robot? Autonomous, automaton self-willed (Greek, auto+matos) Robot Czech writer Karel Capek in 1923 play R.U.R. (Rossum s Universal Robots) labor (Slovak or Polish, robota) workman (Slovak or Polish, robotnik) In the play, robots take over the world!

3 Haptics Robotics from 50,000 feet Task level planning Mapping and Localization HRI Dynamics and Control (ME 646) Mechanisms Motion Planning Vision Manipulation 2/7/14 Applications CS 436/636, G.D. Hager

4 Two Key Problems 1. Motion Planning: Given a mechanism, a known environment, and a known start and goal, compute a set of inputs (e.g. a joint trajectory) that moves the robot from start to goal without collision. 2. Localization and Mapping: Given no a priori knowledge, use information from sensing and motion to simultaneously compute a model of the environment (a map) and robot location within the environment.

5 What is Motion Planning? Determining where to go...more than a search (or it is a geometric search) 2/7/14 CS 436/636, G.D. Hager

6 Basic Path Planning Problem Statement: Compute a continuous sequence of collision-free robot configurations connecting the initial and goal configurations 2D EXAMPLE: Geometry of environment Geometry and kinematics of robot Initial and goal configurations 2/7/14 Path Planner CS 436/636, G.D. Hager Collision-free path

7 Live Motion Planning Experiments Person 1 walks through some obstacles Person 1, looking at Person 2, directs Person 2 through obstacles Person 1, looking at Person 2 with eyes closed, directs Person 2 through obstacles Person 1, looking at a map and not Person 2, whose eyes are still closed, directs Person 2 through obstacles Person 1, looking at an object and Person 2, whose eyes are closed, directs Person 2 to grab an obstacle How do we do the last experiment with a map? 2/7/14 CS 436/636, G.D. Hager

8 What did we assume? Perfect sensors? What information Uncertainty Perfect control? What controls? What else? Uncertainty Perfect thinking? Knowledge of the world? Complete? Processing the world? Everything? 2/7/14 CS 436/636, G.D. Hager

9 Probabilistic Methods/Drug Design Courtesy Lydia Kavraki et al.

10 RoadMap Methods/Animation Path planning in high dimensional spaces Kuffner/Latombe/LaValle

11 Think About Dynamics, Too! Visually controlled juggling 3R manipulator 3-5 full frames / bounce used half-frames Al Rizzi, Noah Cowan, Rob Burridge, & Dan Koditschek estimated single-frame motion obstacle avoidance?!

12 Vision-Based Juggling

13 Planning in Juggling Space

14 SLAM Suppose you wake up in the desert and you need to find your way around --- how do you do it? A Chicken and Egg problem?

15

16

17

18

19

20

21 This Course: We will focus on two robotics problems: Robot Motion Planning: How do I get from A to B in an environment Robot Localization and Mapping: What is the structure of space, and where am I in it. We will primarily focus on algorithms, their analysis, and their implementation (when possible) from real sensor data We will focus on the *static* planning problem the world does not change our own momentum does not matter Almost all of our work will be in simulation although if you take Robocup I/II, you will be able to apply it to real robots

22 Book Principles of Robot Motion: Theory, Algorithms, and Implementations H. Choset, K. M. Lynch, S. Hutchinson, G. Kantor, W. Burgard, L. E. Kavraki and S. Thrun, MIT Press, Boston, /7/14 CS 436/636, G.D. Hager

23 Other Books Planning Algorithms Steven Lavalle, Cambridge University Prress, 2006 Robot Motion Planning, Jean-Claude Latombe, Kluwer, Free download: Probabilistic Robotics S. Thrun, W. Burgard, D. Fox MIT Press, 2006 An Introduction to AI/Robotics Robin Murphy MIT Press, /7/14 CS 436/636, G.D. Hager

24 This Course: Text: Principles of Robot Motion: Theory, Algorithms, and Implementation, Choset et. al, 2005 Chap 1: Introduction Chap 3: Kinematics and Configuration Space Chap 4: Potential Field Methods Chap 5: Roadmaps EXAM Chap 6: Cell decompositions Chap 7: Sampling-Based Motion Planning Chap 8: Kalman-Filter based Localization Chap 9: Bayesian Methods for Localization EXAM

25 Homework Assignments HW 1: Kinematics design HW 2: Simulation HW 3: Probabilistic Roadmap HW 4: SLAM Possibly a final project 2/7/14 CS 436/636, G.D. Hager

26 MoveIt! MoveIt! is a software for mobile manipulation, incorporating the latest advances in motion planning, manipulation, 3D perception, kinematics, control and navigation.

27 General Policies Homework: Roughly 4 assignments (about every 2 weeks) counts for 50% of grade. Exams count the other 50% I expect everyone to come to every lecture I welcome feedback as we go along (I m learning too!)

28 Chapter 1 The Piano-movers problem: Schwarz & Sharir ~ spawned a series of papers on rigid objects moving in space Early work theory based --- Canny PhD planning is Pspacehard! Lozano-Perez (Mason & Taylor) -- notions of configuration space and effective algorithms in the plane (1987) Latombe (early 90 s) -- efficient approximate methods Kavraki (and others, mid/late 90 s) probabilistic methods Now, considered by many to be a solved problem for typical serial chain robots Still some open problems for non-holonomic systems (e.g. a double-semi backing up) and complex dynamics (multiple airliners in turbulance)

29 Robotics is 50! First commercial robot by Unimate sold in /7/14 CS 436/636, G.D. Hager

30 A Brief Pre-History 270 BC: Clocks with movable figures 1495: da Vinci designs mechanical man 1801: Jacquard programmable loom 1921: Capek s RUR 1940: The tortoise 1954: Devol creates first programmable robot 1960: JHU The Beast 2/7/14 CS 436/636, G.D. Hager

31 Classical / Hierarchical Paradigm Sense Plan Act 70s Focus on automated reasoning and knowledge representation STRIPS (Stanford Research Institute Problem Solver): Perfect world model, closed world assumption Find boxes and move them to designated position

32 Shakey 69 Stanford Research Institute

33 Stanford CART Stanford Artificial Intelligence Laboratory / CMU (Moravec)

34 Stanford CART Stanford Artificial Intelligence Laboratory / CMU (Moravec)

35 Historical Search-based AI Shakey the Robot (SRI, ) Triangulating range-finder for sensing obstacles STRIPS based A* planner for navigating to a goal Wireless radio and video camera What have we done since? 2/7/14 CS 436/636, G.D. Hager

36 Blocks World (1960s) Historical Search-based AI 2/7/14 CS 436/636, G.D. Hager

37 Historical Search-based AI STRIPS Action Planning (1960s) 2/7/14 CS 436/636, G.D. Hager

38 Historical Search-based AI STRIPS Action Planning (1960s) 2/7/14 CS 436/636, G.D. Hager

39 Reactive / Behavior-based Paradigm Sense Act No models: The world is its own, best model Easy successes, but also limitations Investigate biological systems

40 Reactive Robots Genghis, Rodney Brooks

41 Polly (Horswill, 1992) The Polly System: specializing perception to a single task/environment a vision-based tour guide at MIT: Polly tested the limits of reactivity.

42 Hybrid Deliberative/reactive Paradigm Plan Sense Act Combines advantages of previous paradigms World model used for planning Closed loop, reactive control

43 Trends in Robotics Research Classical Robotics (mid-70 s) exact models no sensing necessary Hybrids (since 90 s) model-based at higher levels reactive at lower levels Reactive Paradigm (mid-80 s) no models relies heavily on good sensing Probabilistic Robotics (since mid-90 s) seamless integration of models and sensing inaccurate models, inaccurate sensors

44 Robotics Today 2/7/14 CS 436/636, G.D. Hager

45 Incredible Pla-orms for Research Lots of Bodies 1.1. STATE OF THE ART Willow Garage PR2 3 NASA Robonaut Rethink RoboHcs DLR JusHn Willow Garage PR2 Figure 1.1: DLR s Justin ( CS 436/636, G.D. Hager Figure 1.2: Willow Garage s PR2 (

46 Robots in Space Opportunity & Spirit on Mars since Jan. 04

47 In the Air By 2020, all aircraft bought by the military will be unmanned

48 Land and Sea... RHex: Koditschek, Rizzi, Cowan,...

49 A New Record hlp:// CS 436/636, G.D. Hager

50 Snake Robots Inspection or Search (courtesy Howie Choset)

51 Scalable robot for dexterous surgery in small spaces Nabil Simaan, Russell Taylor, Paul Flint, MD 4.2 mm diameter >1 N force application ability -- Impedance control master and slave manipulator -- Remote wrist for high dexterity -- Stereo endoscopic visualization

52 Applications: Medicine Courtesy Peter Kazanzides -- Integrated Surgical Systems, Inc

53 Intuitive Surgical da Vinci System K v Master Position Slave Position -- Impedance control master and slave manipulator -- Remote wrist for high dexterity -- Stereo endoscopic visualization

54 Heart Lander Riviere/Patronik/Zenati (CMU/U. Pittsburgh) Subxiphoid approach attach to beating heart via suction maneuver to site perform therapy Working Channel A new concept for minimally invasive surgery!

55 How Do Robots Today Make a Living? Courtesy KUKA RoboHcs Courtesy of IntuiHve Surgical ROOMBA, Courtesy IRobot KIVA Systems CS 436/636, G.D. Hager

56 Humanoids (DARPA Robotics Challenge) CMU Tartan 2/7/14 NASA Valkyrie CS 436/636, G.D. Hager

57 Minerva: A Museum Tour Guide CMU/University of Bonn

58 Your Assignment3 Read chapter 3 for the next class!