Due Tuesday, October 30, at 12:25, either in class or on Canvas.
Use the rti_stub.py or rti_stub.m file to test my multichannel RTI and VRTI code. Use it with the collected data file HW3_2.txt and HW3_3.txt. Note when you download the Python code, you need to remove the .txt at the end of the filename. The text data files are in a zip files, and the Matlab .m files are also in a zip, so uncompress them.
- Turn the script into a function that can be called, with the input arguments to the function being one or more of the parameters (e.g., excessPathLength, delta, or buffL), and the output being the average error. Then, write a script with a for loop that goes through a list of possible parameter values, records the average error for each parameter value, and then at the end of the for loop, finds the parameter value that minimizes the RMSE. For this purpose, turn off the plotting so that the script runs faster. Have at least 20 values (or the max possible, in the case that 20 is not possible) in the list of possible parameter values. Turn in a plot of the RMSE vs. parameter value. Do this for three different parameters. When running the 2nd test, use the optimal parameter value for the first parameter, etc.
Modify the method. For example:
- You might modify the way the multiple channels are combined
- You might combine variance and shadowing in some way
- You might combine images from two different methods in some way
- You might use a different score vector (eg, your line crossing score from HW2) instead of shadowing or variance
- You might change the weight model (i.e., the weight matrix W in the initRTI function)
- You might add tracking to the coordinate estimate
Try three modifications, and test the results. Describe your modification (including code segments if needed for explanation) and compare its results to those of the original method.