summaryrefslogtreecommitdiffstats
path: root/uav.m
diff options
context:
space:
mode:
Diffstat (limited to 'uav.m')
-rw-r--r--uav.m187
1 files changed, 136 insertions, 51 deletions
diff --git a/uav.m b/uav.m
index 62f6543..ca33346 100644
--- a/uav.m
+++ b/uav.m
@@ -9,11 +9,15 @@
clear; clc; close all; s = tf('s');
% Flags to speed up running for debugging
-do_plots = true;
+do_plots = true; % runs faster without
do_lqr = false; % unused
-do_hinf = true; % midterm
+do_hinf = false; % midterm
do_musyn = true; % endterm
+if do_hinf & do_musyn
+ error('Cannot do both H-infinity and mu synthesis.')
+end
+
fprintf('Controller synthesis for ducted fan VTOL micro-UAV\n')
fprintf('Will do:\n')
if do_plots
@@ -41,15 +45,21 @@ params = uav_params();
% ------------------------------------------------------------------------
%% Define performance requirements
-if do_hinf | do_musyn
+if do_hinf
fprintf('Generating performance requirements...\n')
- perf = uav_performance(params, do_plots);
+ perf = uav_performance_hinf(params, do_plots);
+end
+if do_musyn
+ fprintf('Generating performance requirements...\n')
+ perf = uav_performance_musyn(params, do_plots);
end
% ------------------------------------------------------------------------
%% Define stability requirements
-if do_musyn
+% Note: for hinf it is needed to call uav_mode, but hinf will not actually
+% make use of this struct
+if do_hinf | do_musyn
fprintf('Generating stability requirements...\n')
uncert = uav_uncertainty(params, do_plots);
end
@@ -119,33 +129,48 @@ if do_musyn
fprintf('Performing mu-synthesis controller design...\n')
idx = model.uncertain.index;
- P = model.uncertain.StateSpace([idx.OutputUncertain; idx.OutputError; idx.OutputNominal], ...
- [idx.InputUncertain; idx.InputExogenous; idx.InputNominal]);
-
- % Initial values for D-K iteration
- D_left = eye(model.uncertain.Nz + model.uncertain.Ne + model.uncertain.Ny);
- D_right = eye(model.uncertain.Nv + model.uncertain.Nw + model.uncertain.Nu);
+ P = minreal(model.uncertain.StateSpace(...
+ [idx.OutputUncertain; idx.OutputError; idx.OutputNominal], ...
+ [idx.InputUncertain; idx.InputExogenous; idx.InputNominal]), ...
+ [], false);
% Options for H-infinity
nmeas = model.uncertain.Ny;
nctrl = model.uncertain.Nu;
- hinfopt = hinfsynOptions('Display', 'on', 'Method', 'RIC', ...
- 'AutoScale', 'off', 'RelTol', 1e-3);
+ hinfopt = hinfsynOptions('Display', 'off', 'Method', 'RIC', ...
+ 'AutoScale', 'on', 'RelTol', 1e-2);
% Number of D-K iterations
- niters = 4;
+ niters = 8;
% Frequency raster resolution to fit D scales
- nsamples = 51;
- omega = logspace(-2, 2, nsamples);
+ nsamples = 61;
+ omega = logspace(-2, 3, nsamples);
+
+ % Initial values for D-K iteration
+ nleft = model.uncertain.Nz + model.uncertain.Ne + model.uncertain.Ny;
+ nleft_clp = model.uncertain.Nz + model.uncertain.Ne;
+
+ nright = model.uncertain.Nv + model.uncertain.Nw + model.uncertain.Nu;
+ nright_clp = model.uncertain.Nv + model.uncertain.Nw;
+
+ D_left = tf(eye(nleft));
+ D_right = tf(eye(nright));
+
+ last_mu_rp = inf;
+ mu_plot_legend = {};
% Start DK-iteration
for it = 1:niters
fprintf(' - Running D-K iteration %d ...\n', it);
% Find controller using H-infinity
- [K, ~, gamma, ~] = hinfsyn(D_left * P * D_right, nmeas, nctrl);
+ [K, ~, gamma, ~] = hinfsyn(D_left * P * D_right, nmeas, nctrl, hinfopt);
fprintf(' H-infinity synthesis gamma: %g\n', gamma);
+ if gamma == inf
+ fprintf(' Failed to synethesize H-infinity controller\n');
+ break;
+ end
% Calculate frequency response of closed loop
N = minreal(lft(P, K), [], false); % slient
@@ -156,43 +181,103 @@ if do_musyn
mu_rp = norm(mu_bounds(1,1), inf, 1e-6);
fprintf(' Mu value for RP: %g\n', mu_rp)
+ if do_plots
+ fprintf(' Plotting mu\n');
+ figure(100); hold on;
+ bodemag(mu_bounds(1,1));
+ mu_plot_legend = {mu_plot_legend{:}, sprintf('$\\mu_{%d}$', it)};
+ title('\bfseries $\mu_\Delta(\omega)$ for both Stability and Performance', 'interpreter', 'latex');
+ legend(mu_plot_legend, 'interpreter', 'latex');
+ grid on;
+ drawnow;
+ end
+
+ % Are we done yet?
+ if mu_rp < 1
+ fprintf(' - Found robust controller that meets performance.\n');
+ break
+ end
+
% Fit D-scales
- [dsysl, dsysr] = mussvunwrap(mu_info);
- D_left = fitfrd(genphase(dsysl(1,1)), 1);
- D_right = fitfrd(genphase(dsysr(1,1)), 1);
+ % There are three complex, square, full block uncertainties and
+ % a non-square full complex block for performance
+ [D_left_samples, D_right_samples] = mussvunwrap(mu_info);
+
+ % D scale for alpha uncertainty (first block)
+ i = 1;
+ D_left_samples_alpha = D_left_samples(i, i);
+ D_alpha = fitmagfrd(D_left_samples_alpha, 2);
+
+ % D scale for omega uncertainty (second block)
+ i = model.uncertain.BlockStructure(1, 1) + 1; % after first block
+ D_left_samples_omega = frd(D_left_samples(i, i));
+ D_omega = fitmagfrd(D_left_samples_omega, 3);
+
+ % D scale for state uncertainty (third block)
+ i = model.uncertain.BlockStructure(2, 1) + 1; % after second block
+ D_left_samples_state = D_left_samples(i, i);
+ D_state = fitmagfrd(D_left_samples_state, 5);
+
+ % D scale for performance (non-square)
+ i = model.uncertain.BlockStructurePerf(3, 1); % after third block
+ D_left_samples_perf = D_left_samples(i, i);
+ D_perf = fitmagfrd(D_left_samples_perf, 2);
+
+ % Construct full matrices
+ D_right = blkdiag(D_alpha * eye(4), ...
+ D_omega * eye(1), ...
+ D_state * eye(12), ...
+ D_perf * eye(10), ...
+ eye(5));
+
+ D_left = blkdiag(D_alpha * eye(4), ...
+ D_omega * eye(1), ...
+ D_state * eye(12), ...
+ D_perf * eye(14), ...
+ eye(12));
+
+ % Plot fitted D-scales
+ if do_plots
+ fprintf(' Plotting D-scales ');
+ f = figure(101); clf(f); hold on;
+
+ bodemag(D_left_samples_alpha, omega);
+ bodemag(D_alpha, omega);
+ fprintf('.');
+
+ bodemag(D_left_samples_omega, omega);
+ bodemag(D_omega, omega);
+ fprintf('.');
+
+ bodemag(D_left_samples_state, omega);
+ bodemag(D_state, omega);
+ fprintf('.');
+
+ bodemag(D_left_samples_perf, omega);
+ bodemag(D_perf, omega);
+ fprintf('.');
+
+ fprintf('\n');
+ title(sprintf('\bfseries $D(\\omega)$ Scales Approximations at Iteration %d', it), ...
+ 'interpreter', 'latex')
+ legend(...
+ '$D_{\alpha}$', '$\hat{D}_{\alpha}$', ...
+ '$D_{\omega}$', '$\hat{D}_{\omega}$', ...
+ '$D_{\mathbf{x}}$', '$\hat{D}_{\mathbf{x}}$', ...
+ '$D_{\Delta}$', '$\hat{D}_{\Delta}$', ...
+ 'interpreter', 'latex' ...
+ );
+ grid on;
+ drawnow;
+ end
+ end
+
+ if mu_rp > 1
+ fprintf(' - Failed to synthesize robust controller that meets the desired performance.\n');
+ else
+ ctrl.musyn = struct('K', K, 'mu', mu_rp);
end
end
% ------------------------------------------------------------------------
%% Verify performance satisfaction via mu-analysis
-
-% omega = logspace(-3, 3, 250);
-%
-% N_inf = lft(P, K_inf);
-% N_inf_frd = frd(N_inf, omega);
-%
-% % robust stability
-% [mu_inf_rs, ~] = mussv(N_inf_frd(idx.OutputUncertain, idx.InputUncertain), ...
-% model.uncertain.BlockStructure);
-%
-% % robust performance
-% blk_perf = [model.uncertain.BlockStructure;
-% model.uncertain.BlockStructurePerf];
-%
-% [mu_inf_rp, ~] = mussv(N_inf_frd, blk_perf);
-%
-% % nominal performance
-% mu_inf_np = svd(N_inf_frd(idx.OutputError, idx.InputExogenous));
-%
-% if do_plots
-% figure; hold on;
-% bodemag(mu_inf_rs(1));
-% bodemag(mu_inf_np(1));
-% bodemag(mu_inf_rs(2));
-% bodemag(mu_inf_np(2));
-%
-% grid on;
-% title('$\mu_\Delta(N)$', 'Interpreter', 'latex');
-% end
-
-% vim: ts=2 sw=2 et: