ROBOTICS AND AUTONOMOUS SYSTEMS

Transcription

1 ROBOTICS AND AUTONOMOUS SYSTEMS Simon Parsons Department of Computer Science University of Liverpool LECTURE 6

2 PERCEPTION/ODOMETRY

3 comp parsons-lect06 2/43 Today We ll talk about perception and motor control.

4 comp parsons-lect06 3/43 Perception Sensors give important feedback from the environment Without them, robots are blind. percepts Environment sensors effectors actions Perception is all about what can be sensed and what we can do with that sensing.

5 comp parsons-lect06 4/43 Proprioceptive sensors Classification of sensors Measure values internally to the system (robot), (motor speed, wheel load, heading of the robot, battery status) Exteroceptive sensors Information from the robots environment (distances to objects, intensity of the ambient light, unique features.) Passive sensors Energy coming from the environment. Active sensors Emit their own energy and measure the reaction Better performance, but some influence on envrionment

6 comp parsons-lect06 5/43 Proprioceptive sensors Classification of sensors Measure values internally to the system (robot), (motor speed, wheel load, heading of the robot, battery status) Exteroceptive sensors Information from the robots environment (distances to objects, intensity of the ambient light, unique features.) Passive sensors Energy coming from the environment. Active sensors Emit their own energy and measure the reaction Better performance, but some influence on envrionment

7 comp parsons-lect06 6/43 Proprioceptive sensors Classification of sensors Measure values internally to the system (robot), (motor speed, wheel load, heading of the robot, battery status) Exteroceptive sensors Information from the robot s environment (distances to objects, intensity of the ambient light, unique features) Passive sensors Energy coming from the environment. Active sensors Emit their own energy and measure the reaction Better performance, but some influence on envrionment

8 comp parsons-lect06 7/43 Wheel encoders Measure position or speed of the wheels or steering. Wheel movements can be integrated to get an estimate of the robot s position Odometry Optical encoders are proprioceptive sensors Position estimate is only useful for short movements. Typical resolutions: 2000 increments per revolution.

9 comp parsons-lect06 8/43 Count the changes from black to white: Measure light passing through the encoder. Bounce light off the encoder. Simple encoder will give you count/speed. Quadrature encoder will you direction also. Look at phase of signals from the two bands on the encoder.

10 comp parsons-lect06 9/43 Can obtain additional information (Tom Lackamp)

11 comp parsons-lect06 10/43 Differential Pilot Distance information on its own permits a crude form of navigation: Dead reckoning Calculate how far the robot has gone based on wheel rotations. LeJOS provides some of this in the: class DifferentialPilot Though our robot doesn t have a differential drive, this class sort of works with it.

12 comp parsons-lect06 11/43 Constructor: DifferentialPilot( double wheeldiameter, double trackwidth, RegulatedMotor leftmotor, RegulatedMotor rightmotor) wheeldiameter is in any units you like But then you have to use the same units for other Pilot commands. trackwidth is the distance between the left and right wheels.

13 comp parsons-lect06 12/43 Set speed of motion settravelspeed(double speed) speed is in wheel-diameters-units per second. Move a certain amount travel(double distance) distance is in wheel-diameters-units. Rotate: rotate(double angle) rotate through specified angle (in degrees) in a zero-radius turn. Lots of other methods defined in the API.

14 comp parsons-lect06 13/43 From the way it handles wheel diameter, it looks like DifferentialPilot just counts wheel rotations: Since: revolutions = distance π diameter

15 comp parsons-lect06 14/43 A example program import lejos.nxt.*; import lejos.robotics.navigation.differentialpilot; public class SimplePilot{ DifferentialPilot pilot ; public void drawsquare(float length){ for(int i = 0; i<4 ; i++){ pilot.travel(length); // Drive forward pilot.rotate(90); // Turn 90 degrees } } } public static void main(string[] args){ SimplePilot sp = new SimplePilot(); sp.pilot = new DifferentialPilot( 3.25, 19, Motor.B, Motor.C); sp.drawsquare(40); }

16 comp parsons-lect06 15/43 Key lines import lejos.robotics.navigation.differentialpilot; Link the DifferentialPilot class sp.pilot = new DifferentialPilot (3.25, 19, Motor.B, Motor.C); Create a new Pilot object and initialise it. pilot.travel(length); pilot.rotate(90); Call the Pilot to move the robot.

17 comp parsons-lect06 16/43 Calibration When you use DifferentialPilot you will find it doesn t do exactly what you ask it to right away. With robot dimensions Pretty good on distance. Less good on rotation. You will need to callibrate to get it to do what you want it to do. You will need to keep on callibrating.

18 comp parsons-lect06 17/43 Borenstein s experiment

19 comp parsons-lect06 18/43 Borenstein s experiment

20 comp parsons-lect06 19/43 Borenstein s experiment

21 comp parsons-lect06 20/43 Pose Provider When executing a program using the DifferentialPilot, the control loop running the motors knows instantaneously how far the robot has moved. That is what it uses to know when to stop the motors. It is useful to be able to log this information in the control program. The OdometryPoseProvider provides some of this ability.

22 comp parsons-lect06 21/43 Pose Pose objects are manipulated by OdometryPoseProvider. Store a robot pose. Turns out you need to do this a lot. Has no necessary relation to where the robot is. Methods: getx() gety() getheading() Just as you might/should expect. Values returned are floats.

23 comp parsons-lect06 22/43 OPP Methods SetPose(Pose apose) sets the Pose value in the OPP. Note that this does not move the robot, just changes the value that is stored. GetPose() returns a Pose. This is the current pose stored by the OPP. If used correctly, this Pose will tell you something useful.

24 comp parsons-lect06 23/43 OPP Constructor When you create an OPP, you link it to a Pilot object: OdometryPoseProvider opp = new OdometryPoseProvider(dp); where dp is a DifferentialPilot. Then the pose returned by the OPP is updated when the robot moves. Of course, it is updated by the amount that the robot thinks it moves. (Since the robot doesn t actually know how much it is moving.)

25 comp parsons-lect06 24/43 A example program import lejos.nxt.*; import lejos.robotics.navigation.differentialpilot; import lejos.robotics.localization.odometryposeprovider; public class SimplePose { public static void main(string[] args){ DifferentialPilot dp = new DifferentialPilot(3.22, 19, Motor.B, Motor.C); OdometryPoseProvider opp = new OdometryPoseProvider(dp); System.out.println("First pose = " + opp.getpose()); dp.travel(15); System.out.println("Second pose = " + opp.getpose()); } }

26 comp parsons-lect06 25/43 Key lines DifferentialPilot dp = new DifferentialPilot(3.22, 19, Motor.B, Motor.C); OdometryPoseProvider opp = new OdometryPoseProvider(dp); Create the OPP with a reference to the DifferentialPilot System.out.println("First pose = " + opp.getpose()); Access the pose information.

27 comp parsons-lect06 26/43 Heading sensors Heading sensors can be: proprioceptive (gyroscope, inclinometer); or exteroceptive (compass). Used to determine the robot s orientation and/or inclination. Allow, together with an appropriate velocity information, to integrate the movement to an position estimate. A bit more sophisticated than just using odometry.

28 comp parsons-lect06 27/43 Compass Used since before 2000 B.C. Chinese suspended a piece of naturally magnetite from a silk thread and used it to guide a chariot over land. Magnetic field on earth Absolute measure for orientation. Large variety of solutions to measure the earth magnetic field Mechanical magnetic compass Direct measure of the magnetic field (Hall-effect, magnetoresistive sensors)

29 comp parsons-lect06 28/43 Major drawback Weakness of the earth field Easily disturbed by magnetic objects or other sources Not feasible for indoor environments in general. Modern devices can give 3D orientation relative to Earth s magnetic field.

30 comp parsons-lect06 29/43 Gyroscope Heading sensors, that keep the orientation to a fixed frame Provide an absolute measure for the heading of a mobile system. Unlike a compass doesn t measure the outside world. Two categories, mechanical and optical gyroscopes Mechanical Gyroscopes Standard gyro Rate gyro Optical Gyroscopes Rate gyro

31 comp parsons-lect06 30/43 Concept: inertial properties of a fast spinning rotor gyroscopic precession Angular momentum associated with a spinning wheel keeps the axis of the gyroscope inertially stable. No torque can be transmitted from the outer pivot to the wheel axis Spinning axis will therefore be space-stable Quality: 0.1 degrees in 6 hours In rate gyros, gimbals are held by torsional springs. Measuring force gives angular velocity.

32 comp parsons-lect06 31/43 Optical gyroscopes Use two monochromatic light (or laser) beams from the same source. One beam travels clockwise in a cylinder around a fiber, the other counterclockwise. The beam traveling in direction of rotation: Slightly shorter path shows a higher frequency Difference in frequency f of the two beams is proportional to the angular velocity Ω of the cylinder/fiber. Newest optical gyros are solid state.

33 comp parsons-lect06 32/43 Accelerometer Measure acceleration. Mass on a spring. Measure force in spring, gives acceleration of mass. Weight. Any heading or position sensor reading can be differentiated to give acceleration. Difference between two values gives velocity Difference between two velocities gives acceleration Any velocity sensor reading can be handled similarly.

34 comp parsons-lect06 33/43 Sensitivity Sensor performance Ratio of output change to input change In real world environment, the sensor has very often high sensitivity to other environmental changes, e.g. illumination. Cross-sensitivity Sensitivity to environmental parameters that are orthogonal to the target parameters Error / Accuracy Difference between the sensors output and the true value m v accuracy = 1 v where m = measured value and v = true value.

35 comp parsons-lect06 34/43 Sensor performance (more) Systematic error deterministic errors Caused by factors that can (in theory) be modeled prediction e.g. distortion caused by the optics of a camera. Random error non-deterministic No prediction possible However, they can be described probabilistically e.g. error in wheel odometry. Precision Reproducibility of sensor results range σ

36 comp parsons-lect06 35/43 Deterministic sensor errors

37 comp parsons-lect06 36/43 Coping with errors Compensate for systematic errors. Build a probabilistic model of random errors. Obtain a distribution of possible positions.

38 comp parsons-lect06 37/43 A simple error model for straight line motion: (Thrun, Burgard and Fox)

39 comp parsons-lect06 38/43 Which is worse when we turn: (Thrun, Burgard and Fox)

40 comp parsons-lect06 39/43 Coping with errors Compensate for systematic errors. Build a probabilistic model of random errors. Obtain a distribution of possible positions. Know where we are on average.

41 comp parsons-lect06 40/43 Coping with errors Compensate for systematic errors. Build a probabilistic model of random errors. Obtain a distribution of possible positions. Know where we are on average. Don t know where we are in particular

42 comp parsons-lect06 41/43 Coping with errors Compensate for systematic errors. Build a probabilistic model of random errors. Obtain a distribution of possible positions. Know where we are on average. Don t know where we are in particular. Errors accumulate over time.

43 comp parsons-lect06 Possibility of spectacular failure 42/43

44 comp parsons-lect06 43/43 Summary This lecture started to look at sensor data. It concentrated on data that can be used in odometry. Wheel encoders and looked at LeJOS support for doing odometry. Also looked at other kinds of related sensor data: Compass Gyroscope Next time we ll look at range sensor data and cameras as sensors.