Scribd
Upload
41 views12 pages

Assignment 2 Kalma

The document describes using an extended Kalman filter to estimate the position of a robot based on sensor measurements. It involves selecting a Kinect sensor to measure distance and orientation, interfacing it with MATLAB, and using the measurements in an EKF along with derived equations for the Jacobian and covariance matrices. Plots show the EKF providing estimates that follow the true path when measurement noise is added, though the estimates are still not perfectly accurate due to issues like not accounting for theta in the model.

Uploaded by

Javeria Farooq
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
41 views12 pages

Assignment 2 Kalma

The document describes using an extended Kalman filter to estimate the position of a robot based on sensor measurements. It involves selecting a Kinect sensor to measure distance and orientation, interfacing it with MATLAB, and using the measurements in an EKF along with derived equations for the Jacobian and covariance matrices. Plots show the EKF providing estimates that follow the true path when measurement noise is added, though the estimates are still not perfectly accurate due to issues like not accounting for theta in the model.

Uploaded by

Javeria Farooq
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 12

Extended Kalman Filter

Problem Statement
• Place a sensor on the ground that gives you distance and
orientation of the robot with respect to a reference axis.
• Use measurements from sensor along with a Kalman Filter
to estimate [x,y] position of the robot
Sequence
• Selection of Sensor
• Interfacing of Sensor
• Reading laser data
• Determining Jacobian Matrix and Control Input
• MATLAB Programming
• Debugging and Observing noise effects
• Results
Sensor Selection
• Sensor
• Get distance and Orientation of Bob
• Proximity sensors and vision sensors considered in Coopelia Sim
• Kinect Closest Point Calculation Sensor Selected

• Reason for Sensor Selection


• Detect only rendeble objects
• Image content can be accessed via the API
• Code in Lua included in Child Script for reading Coordinates
Sensor Interfacing with MATLAB
• Connectivity observed by accessing VREP library files
• Code written in Child Script to relay data
• Code written in MATLAB to get data
• Noise added in measurements
• Data received successfully
• Plotted measured data without noise to see sensor accuracy
Measurements by Sensor
Steps for Extended Kalman Filter
• Assigning initial values to [X,Y] as [0,0]
• Initiating Covariance Matrixes and Jacobian equal to Identity for process and measurement Noise and taking
initial covariance with a constant determinent (0.01)
• With in the While control loop, calculate apriori covariance
• Use Apriori X and Y (for rst step equals to zero) for calculation of Measured Matrix H(distance and theta)
• Calculate Jacobian using derived formula with predicted theta(theta Temp).
• Calculate Kalman Gain (L) using predicted values and
• Take Measurement (Z) from sensor and convert it into distance and theta.
• Calculate Aposteriori position by considering error between Z and H
• With in while loop, assign calculated position to X and Y.
• Update covariance Matrix
Response of Kalman Without Noise
Response of Kalman With Noise
Extended Kalman Filter with Control inputs
• Assigning initial values to [X,Y] as [0,0]
• Initiating Covariance Matrixes and Jacobian equal to Identity for process and measurement Noise and taking initial
covariance with a constant determinent (0.01)
• With in the While control loop, calculate apriori covariance using Transition function in MATLAB. =Fx and =Fu were
• Calculated as per theory given in page No 271-273 of text book and accordingly error covariance Matrix was calculated
• Use Apriori X and Y (for rst step equals to zero) for calculation of Measured Matrix H(distance and theta)
• Calculate Jacobian using derived formula with predicted theta(theta Temp).
• Calculate Kalman Gain (L) using predicted values and
• Take Measurement (Z) from sensor and convert it into distance and theta.
• Calculate Aposteriori position by considering error between Z and H
• With in while loop, assign calculated position to X and Y.
• Update covariance Matrix
EKF (with Control Input) Response for Noisy
Data
Conclusion
• Not taken into account theta
• Position estimation is still not accurate and is giving what i understood as angle error.
• We have taken initial assumption of covariance equals to a constant value which can be veried to
improve the response of Kalman Filter. Including theta and
• Finding a means to calculate measurement jacobian of theta will also greatly improve the reponse.

You might also like