EE565:Mobile Robotics Lecture 3

Transcription

1 EE565:Mobile Robotics Lecture 3 Welcome Dr. Ahmad Kamal Nasir

2 Today s Objectives Motion Models Velocity based model (Dead-Reckoning) Odometry based model (Wheel Encoders) Sensor Models Beam model of range finders Feature based sensor models Camera Laser scanner Kinect

3 Robot Motion Robot motion is inherently uncertain. How can we model this uncertainty?

4 Dynamic Bayesian Network for Controls, States, and Sensations

5 Probabilistic Motion Models To implement the Bayes Filter, we need the transition model p(x x, u). The term p(x x, u) specifies a posterior probability, that action u carries the robot from x to x. In this section we will specify, how p(x x, u) can be modeled based on the motion equations.

6 Coordinate Systems In general the configuration of a robot can be described by six parameters. Three-dimensional Cartesian coordinates plus three Euler angles pitch, roll, and tilt. Throughout this section, we consider robots operating on a planar surface. The state space of such systems is three-dimensional (x,y,).

7 Typical Motion Models In practice, one often finds two types of motion models: Odometry-based Velocity-based (dead reckoning) Odometry-based models are used when systems are equipped with wheel encoders. Velocity-based models have to be applied when no wheel encoders are given. They calculate the new pose based on the velocities and the time elapsed.

8 Dead Reckoning Derived from deduced reckoning. Mathematical procedure for determining the present location of a vehicle. Achieved by calculating the current pose of the vehicle based on its velocities and the time elapsed.

9 Reasons for Motion Errors ideal case different wheel diameters bump carpet and many more

10 Wheel Encoders A pair of encoders is used on a single shaft. The encoders are aligned so that their two data streams are one quarter cycle (90 deg.) out of phase. Which direction is shaft moving? Suppose the encoders were previously at the position highlighted by the dark band; i.e., Encoder A as 1 and Encoder B as 0. The next time the encoders are checked: If they moved to the position AB=00, the position count is incremented If they moved to the position AB=11, the position count is decremented

11 Odometry Model Robot moves from x, y, to x', y', ' Odometry information u.,, rot1 rot2 trans trans 2 ( x' x) ( y' y atan2( y' y, x' ) rot1 x rot2 ' rot1 ) 2 rot2 x', y', ' x, y, rot1 trans

12 The atan2 Function Extends the inverse tangent and correctly copes with the signs of x and y.

13 Noise Model for Odometry The measured motion is given by the true motion corrupted with noise. ˆ ˆ ˆ rot1 rot1 1 rot1 2 trans trans rot2 rot2 1 rot2 2 trans 3 trans 4 rot1 rot2 trans

14 Typical Distributions for Probabilistic Motion Models Normal distribution Triangular distribution 1 x ( x) e ( x) 0 if x x

15 How to Sample from Normal or Triangular Distributions? Sampling from a normal distribution 1. Algorithm sample_normal_distribution(b): 2. return Sampling from a triangular distribution 1. Algorithm sample_triangular_distribution(b): 2. return

16 Normally/Triangular Distributed Samples 10 6 samples

17 Calculating the Posterior Given x, x, and u 1. Algorithm motion_model_odometry(x,x,u) trans ( x' x) ( y' y) rot 1 atan2( y' y, x' x) rot2 ' rot1 ˆ 2 2 trans ( x' x) ( y' y) ˆ rot 1 atan2( y' y, x' x) ˆ ˆ rot2 ' rot 1 p prob( ˆ, ˆ rot1 rot1 1 rot1 p prob( ˆ, ˆ trans trans 3 p prob( ˆ, ˆ 1 ˆ 2 trans ) ( ˆ 4 rot1 ˆ ) 2 trans rot2 3 rot2 rot2 1 rot2 2 trans 11. return p 1 p 2 p 3 odometry values (u) values of interest (x,x ) ˆ ))

18 Application Repeated application of the sensor model for short movements. Typical banana-shaped distributions obtained for 2dprojection of 3d posterior. p(x u,x ) x u x u

19 Examples (Odometry-Based)

20 Velocity-Based Model

21 Posterior Probability for Velocity Model

22 Examples (velocity based)

23 Map-Consistent Motion Model p( x u, x') p( x u, x', m) Approximation: p( x u, x', m) p( x m) p( x u, x')

24 Sensors for Mobile Robots Contact sensors: Bumpers Internal sensors Accelerometers (spring-mounted masses) Gyroscopes (spinning mass, laser light) Compasses, inclinometers (earth magnetic field, gravity) Proximity sensors Sonar (time of flight) Radar (phase and frequency) Laser range-finders (triangulation, tof, phase) Infrared (intensity) Visual sensors: Cameras Satellite-based sensors: GPS

25 Proximity Sensors The central task is to determine P(z x), i.e., the probability of a measurement z given that the robot is at position x. Question: Where do the probabilities come from? Approach: Let s try to explain a measurement.

26 Beam-based Sensor Model Scan z consists of K measurements. z z, z,..., z } { 1 2 K Individual measurements are independent given the robot position. P( z x, m) K k1 P( z k x, m)

27 Beam-based Sensor Model P( z x, m) K k1 P( z k x, m)

28 Typical Measurement Errors of an Range Measurements 1. Beams reflected by obstacles 2. Beams reflected by persons / caused by crosstalk 3. Random measurements 4. Maximum range measurements

29 Proximity Measurement Measurement can be caused by a known obstacle. cross-talk. an unexpected obstacle (people, furniture, ). missing all obstacles (total reflection, glass, ). Noise is due to uncertainty in measuring distance to known obstacle. in position of known obstacles. in position of additional obstacles. whether obstacle is missed.

30 Beam-based Proximity Model Measurement noise Unexpected obstacles 0 z exp z max 0 z exp z max P hit ( z x, m) 1 2b e 1 ( zz 2 b 2 exp ) P unexp ( z x, m) e 0 z z z exp otherwise

31 Beam-based Proximity Model Random measurement Max range 0 z exp z max 0 z exp z max P rand ( z 1 x, m) z max P max ( z x, m) 1 z small

32 Resulting Mixture Density ), ( ), ( ), ( ), ( ), ( rand max unexp hit rand max unexp hit m x z P m x z P m x z P m x z P m x z P T How can we determine the model parameters?

33 Approximation Results Laser Sonar 300cm 400cm

34 Approximation Results Laser Sonar

35 Example z P(z x,m)

36 San Jose Tech Museum Occupancy grid map Likelihood field

37 Landmarks Active beacons (e.g., radio, GPS) Passive (e.g., visual, retro-reflective) Standard approach is triangulation Sensor provides distance, or bearing, or distance and bearing.

38 Distance and Bearing

39 Probabilistic Model 2 2 ) ) ( ( ) ) ( ( ˆ y i m x i m d y x ), prob( ˆ ), prob( ˆ det d d d p,,,,, y x x d i z ) ) (, ) ( atan2( ˆ x i m y i m a x y ), uniform ( fp det det m x z P z p z

40 Distributions

41 Laser Scanner Features (Line) n n α = 1 2 atan2( 2 y y i x x i, i=0 i=0 r = x cos α + y sin α y y i 2 x x i 2 ) ρ cos θ α r = 0 P αr = σ α 2 σ αr σ rα σ r 2 σ α 2 = σ r 2 = σ αr = σ rα = n i=0 n i=0 n i=0 α ρ i r ρ i 2 2 σ2 ρi σ2 ρi α r ρ i ρ i 2 σ ρi 01 Feb 2016 Dr. -Ing. Ahmad Kamal Nasir

42 2D Geometric Feature based Map

43 Kinect Features (Plane) p x sin θ cos φ + p y sin θ sin φ + p z cos θ = ρ Do until DetectedPlanes < 8 and TotalPoints > MinPoints Randomly select (P 1, P 2, P 3 ) from a random circular region Calculate plane from (P 1, P 2, P 3 ) Transform the Calculate plane from R 3 to Hough space If local maxima is found in Hough space Delete points corresponding to plane from input points Calculate plane boundries Reset Hough space End if End Do

44 3D Geometric Feature based Map

45 Summary of Sensor Models Explicitly modeling uncertainty in sensing is key to robustness. In many cases, good models can be found by the following approach: 1. Determine parametric model of noise free measurement. 2. Analyze sources of noise. 3. Add adequate noise to parameters (eventually mix in densities for noise). 4. Learn (and verify) parameters by fitting model to data. 5. Likelihood of measurement is given by probabilistically comparing the actual with the expected measurement. This holds for motion models as well. It is extremely important to be aware of the underlying assumptions!

46 Summary Motion Models Velocity based model (Dead-Reckoning) Odometry based model (Wheel Encoders) Sensor Models Beam model of range finders Feature based sensor models Camera Laser scanner Kinect

47 Questions