Intelligent and Adaptive Systems
Budget: $10 – $30 USD
1) You are supplied with a solution of the 2R planar manipulator IK using the ANFIS tool from the MATLAB Fuzzy Logic Toolbox. You are required to extend this to model the IK of a 3R planar manipulator with the following geometry:
Link Lengths:
Link Length
1 7
2 5
3 3
With all angles being limited from 0 to ( π)/2
Explain the methods used to extend the model.
2) Train the system with a known closed-form forwards kinematics solution at a fixed number of positions and then for ANFIS to provide approximate solutions to the IK over the whole usable workspace of the manipulator. Then demonstrate the results obtained through calculations, coding, and experiments.
3) You will need to experiment with a number of parameters to find a system which learns with sufficient accuracy but without unacceptably long training times. It is particularly important to demonstrate that the trained network generalises well using suitable validation methods. Explain validation testing you used including mathematical formulas and critical assess the performance of your system.
4) Design an MLPL Neural Network to achieve the task of the system you designed in 1.
5) Compare the ANFIS model with MLP Neural Network in terms of performance including (error, training time, size of network, etc).
Use Proper references, layout/presentation.
Reporting
You have to write a report, using not more than 2500 words to describe your investigations and results.
You need to:
1) Demonstrate that you understand the theory behind the approaches you use to solve a problem.
2) Explain the methods used and describe the steps taken to validate your results.
3) Include critical assessment and analysis of the relative merits of the approaches you have used.
4) Include Discussion of Results and Conclusions.
5) Provide references using the Harvard system
http://www1.uwe.ac.uk/students/studysupport/studyskills/referencing/uweharvard.aspx
6) Provide any code you have written in an appendix. (This will not be included in the word count)
Link Lengths:
Link Length
1 7
2 5
3 3
With all angles being limited from 0 to ( π)/2
Explain the methods used to extend the model.
2) Train the system with a known closed-form forwards kinematics solution at a fixed number of positions and then for ANFIS to provide approximate solutions to the IK over the whole usable workspace of the manipulator. Then demonstrate the results obtained through calculations, coding, and experiments.
3) You will need to experiment with a number of parameters to find a system which learns with sufficient accuracy but without unacceptably long training times. It is particularly important to demonstrate that the trained network generalises well using suitable validation methods. Explain validation testing you used including mathematical formulas and critical assess the performance of your system.
4) Design an MLPL Neural Network to achieve the task of the system you designed in 1.
5) Compare the ANFIS model with MLP Neural Network in terms of performance including (error, training time, size of network, etc).
Use Proper references, layout/presentation.
Reporting
You have to write a report, using not more than 2500 words to describe your investigations and results.
You need to:
1) Demonstrate that you understand the theory behind the approaches you use to solve a problem.
2) Explain the methods used and describe the steps taken to validate your results.
3) Include critical assessment and analysis of the relative merits of the approaches you have used.
4) Include Discussion of Results and Conclusions.
5) Provide references using the Harvard system
http://www1.uwe.ac.uk/students/studysupport/studyskills/referencing/uweharvard.aspx
6) Provide any code you have written in an appendix. (This will not be included in the word count)