I am trying to implement a particle filter and run it on saved data. One function "pf2_1" loops through the data and uses a particle filter constructed in the class "Model". It is in a separate class for organization purposes.
When I try and construct Model, I get the following error:
Not enough input arguments.Error in pf2_1 (line 52) model = Model(sensorPositions, gsd_, np_, psf_, rsm_);I added underscores after those inputs for fear I was having a name conflict. No matter what I try I cannot get past the construction. Below are my files:
pf2_1:
% D is [secs, mins, raw data... raw data, state est 1... state est n]
%function [avgDist, avgAng, stddevDist, stddevAng, D] = doPF(dataFile, aggFname, directory, params)
function [avgDistMean, avgAngMean, stddevDistMean, stddevAngMean, avgDistMax, avgAngMax, stddevDistMax, stddevAngMax, D] = pf2_1(simFile, makePlot, gsd_, wbd, psf_, np_, sem, rsm_) %clear all
%close all
%fclose('all');
delete(instrfindall) % For open serial ports
% Output file
outputname = ''; % Declare # of receivers
NUM_RECEIVERS = 6; sensorPositions = zeros(2, NUM_RECEIVERS); % Plot or no plot?
PLOT = makePlot; myPlot = []; % Live or simulated?
LIVE = false; if (LIVE) % Specify beacon locations
% initialize beacons (do all this in beacon-handling class?)
% collect garbage
% ENSURE ALL OF THEM ARE TALKING TO YOU
% They are talking? Good, start up everything else
% Create file to log to (inside file handling class!!)
else sim_file = simFile; data = csvread(sim_file); %disp(size(sensorPositions(1,:)));
%disp(size(data(1,2:2+NUM_RECEIVERS)));
sensorPositions(1,:) = data(1,3:2+NUM_RECEIVERS); sensorPositions(2,:) = data(2,3:2+NUM_RECEIVERS); %disp(sensorPositions);
% Create model (includes particle filter, mapping RSSI-m, etc...
% ====================== ERROR OCCURS HERE ======================
model = Model(sensorPositions, gsd_, np_, psf_, rsm_); %model = Model();
%model.begin(sensorPositions, gsd_, np_, psf_, rsm_);
% Create plot class
if (PLOT) myPlot = PlotClass(); myPlot.begin(max(sensorPositions(1,:)), max(sensorPositions(2,:)), ... sensorPositions, model.bound, ... LIVE); % Not live if in this branch, leave for ease of reuse
end disp('starting loop'); distMeanList = []; % Can predetermine size for improved speed
angleMeanList = []; distMaxList = []; angleMaxList = []; D = []; % Loop over data
for s=3:length(data) [statePred, covPred] = predict(model.pf, model.noise); %disp('exited predict');
secs = data(s, 1); mins = data(s, 2); raws = data(s, 3:2+NUM_RECEIVERS); x_truth = data(s, 3+NUM_RECEIVERS); y_truth = data(s, 4+NUM_RECEIVERS); %mapped = model.RSSI_TO_M_COEFF * exp(model.RSSI_TO_M_EXP * data(s, 3:2+NUM_RECEIVERS));
mapped = model.RSSI_TO_M_COEFF * exp(model.RSSI_TO_M_EXP * raws); %[stateCorrected, covCorrected] = correct(model.pf, data(s, 3:2+NUM_RECEIVERS), sensorPositions);
[stateCorrected, covCorrected] = correct(model.pf, mapped, sensorPositions); %disp('exited correct');
%disp(stateCorrected);
% Get dist from truth
diff = sqrt((stateCorrected(1)-y_truth)^2 + (stateCorrected(2)-y_truth)^2); %disp(data(s, 3+NUM_RECEIVERS:4+NUM_RECEIVERS));
%disp(stateCorrected); angle = atan2(x_truth-stateCorrected(1), y_truth-stateCorrected(2)); D = [D, [stateCorrected, covCorrected]]; distMeanList = [distMeanList, diff]; angleMeanList = [angleMeanList, angle]; notNeeded, idx = max(model.pf.Weights); diffMax = sqrt((model.pf.Particles(idx,2)-y_truth)^2 + (model.pf.Particles(idx,1)-y_truth)^2); angleMax = atan2(x_truth-model.pf.Particles(idx,1), y_truth-model.pf.Particles(idx,2)); distMaxList = [distMaxList, diffMax]; angleMaxList = [angleMaxList, angleMax]; %disp(diff);
%disp(angle);
%disp(sprintf('%f, %f', diff, angle));
%rawString = sprintf('%d, ', raws);
% secs, mins, raw data, x truth, y truth, x meas, y meas, diff, angle
%toPrint = sprintf('%f,%f,%s,%d,%d,%d,%d,%f,%f', secs, mins, rawString, x_truth, y_truth, stateCorrected(1), stateCorrected(2), diff, angle);
%disp(toPrint);
%fprintf(fid, '%s', toPrint);
%fprintf(fid, '\n');
if (PLOT) %myPlot.updatePlotSim(model.pf, data(s,1), stateCorrected, data(s, 2+NUM_RECEIVERS:3+NUM_RECEIVERS));
myPlot.updatePlotSim(model.pf, data(s,1), mapped, sensorPositions, stateCorrected, [x_truth, y_truth]); disp('updated plot'); end end avgDistMean = mean(distMeanList); avgAngMean = mean(angleMeanList); stddevDistMean = std2(distMeanList); stddevAngMean = std2(angleMeanList); avgDistMax = mean(distMaxList); avgAngMax = mean(angleMaxList); stddevDistMax = std2(distMaxList); stddevAngMax = std2(angleMaxList); endendModel:
classdef Model < handle% Holds all the tuable parameters of the particle filter
properties(SetAccess = public) % The particle filter
%pf = robotics.ParticleFilter % Is this allowed here?
pf % # of particles in the filter
NUM_PARTICLES = 1000 % Boundaries of the system, square with edge 2*bound
bound = 10 % sensor positions (x, y)
sensorPositions % Noise parameter
noise = 0.1 % State estimation method
sem = 'mean' % or 'maxweight'
% Resampling mehtod 'multinomial' or 'systematic'
rsm = 'systematic' % Gaussian stddev in mlf
gsd % RSSI to m conversion coeffs
RSSI_TO_M_COEFF = 0.00482998 RSSI_TO_M_EXP = -0.104954 end properties(SetAccess = private) end methods% function obj = Model(sensorPositions, gsd, numParticles, psf, rsm)
%function obj = Model()
% %pf = robotics.ParticleFilter; % Do stuff like this in the
% %properties
% obj.pf = robotics.ParticleFilter;
%
% obj.gsd = gsd;
% obj.NUM_PARTICLES = numParticles;
% obj.noise = psf;
% %obj.sem = sem;
% obj.rsm = rsm;
% % stateBounds = [
% -obj.bound, obj.bound;
% -obj.bound, obj.bound];
% initialize(obj.pf, obj.NUM_PARTICLES, stateBounds);
% % obj.pf.StateEstimationMethod = 'mean'; % is this allowed in the properties?
% obj.pf.ResamplingMethod = obj.rsm;
% % obj.pf.StateTransitionFcn = @stf1_1;
% obj.pf.MeasurementLikelihoodFcn = @mlf1_1;
% % obj.sensorPositions = sensorPositions;
% Do nothing?
%end
%function begin(obj, array)
%function begin(obj, sensorPosition, gsd, numParticles, psf, rsm)
function obj = Model(sensorPosition_, gsd_, numParticles_, psf_, rsm_) %[sensorPosition, gsd, numParticles, psf, rsm] = array;
% sensorPosition = array(0);
% gsd = array(1);
% numParticles = array(2);
% psf = array(3);
% rsm = array(4);
obj.pf = robotics.ParticleFilter; obj.gsd = gsd_; obj.NUM_PARTICLES = numParticles_; obj.noise = psf_; %obj.sem = sem;
obj.rsm = rsm_; stateBounds = [ -obj.bound, obj.bound; -obj.bound, obj.bound]; initialize(obj.pf, obj.NUM_PARTICLES, stateBounds); obj.pf.StateEstimationMethod = 'mean'; % is this allowed in the properties?
obj.pf.ResamplingMethod = obj.rsm; obj.pf.StateTransitionFcn = @stf1_1; obj.pf.MeasurementLikelihoodFcn = @mlf1_1; obj.sensorPositions = sensorPosition_; end function [stateCorrected, covCorrected] = correct(pf, measurement, sensorPositions) [stateCorrected, covCorrected] = correct(pf, measurement, sensorPositions, obj.gsd); end function [statePred, covPred] = predict(pf, NOISE) [statePred, covPred] = predict(pf, NOISE); end endendAs you can see I have also tried an initialization function for Model (begin) instead of doing everything in the constructor and received the same error. I also tried passing the variables in an array instead of one at a time and received the same "Not enough inputs" error.
If you see any other ways I can improve my organization, efficiency, etc your suggestions are appreciated.
Best Answer