1function [ph, phases] = refreshProcessRepresentations(self)
2% [PH, PHASES] = REFRESHPROCESSREPRESENTATIONS()
4% Copyright (c) 2012-2026, Imperial College London
8M = getNumberOfStations(self);
9K = getNumberOfClasses(self);
12 ph{ist,1} = cell(1,K);
15stations = self.stations;
17 if ist == self.getIndexSourceStation
18 ph_i = stations{ist}.getSourceRates();
20 switch class(stations{ist})
28 ph_i = Coxian(mu_i,phi_i).getProcess;
35 ph_i{r} = Coxian(mu_i{r},phi_i{r}).getProcess;
38 ph_i = stations{ist}.getServiceRates();
43 % NHPP carries a rate schedule, not D0/D1: always 1 active phase
44 isSchedule_ir =
false;
45 if isa(stations{ist},
'Source') && any(ist == self.getIndexSourceStation) ...
46 && length(stations{ist}.input.sourceClasses) >= r ...
47 && ~isempty(stations{ist}.input.sourceClasses{r})
48 isSchedule_ir = ismethod(stations{ist}.input.sourceClasses{r}{end},
'getRateSchedule');
49 elseif isa(stations{ist},
'ServiceStation') ...
50 && length(stations{ist}.server.serviceProcess) >= r ...
51 && ~isempty(stations{ist}.server.serviceProcess{r})
52 isSchedule_ir = ismethod(stations{ist}.server.serviceProcess{r}{end},
'getRateSchedule');
54 if isempty(ph{ist}{r}) % fluid fails otherwise
58 elseif ~isMAP(ph{ist}{r})
59 % Non-Markovian distribution: convert to MAP representation
60 ph{ist}{r} = convertToMAP(stations{ist}, r, ph{ist}{r});
61 phases(ist,r) = length(ph{ist}{r}{1});
62 elseif any(isnan(ph{ist}{r}{1}(:))) || any(isnan(ph{ist}{r}{2}(:))) % disabled
65 phases(ist,r) = length(ph{ist}{r}{1});
69if ~isempty(self.sn) %&& isprop(self.sn,
'mu')
75 % NHPP carries a rate schedule, not D0/D1: skip map_pie
76 isSchedule_ir =
false;
77 if isa(stations{ist},
'Source') && any(ist == self.getIndexSourceStation) ...
78 && length(stations{ist}.input.sourceClasses) >= r ...
79 && ~isempty(stations{ist}.input.sourceClasses{r})
80 isSchedule_ir = ismethod(stations{ist}.input.sourceClasses{r}{end},
'getRateSchedule');
81 elseif isa(stations{ist},
'ServiceStation') ...
82 && length(stations{ist}.server.serviceProcess) >= r ...
83 && ~isempty(stations{ist}.server.serviceProcess{r})
84 isSchedule_ir = ismethod(stations{ist}.server.serviceProcess{r}{end},
'getRateSchedule');
87 proc{ist}{r} = map_ir;
91 pie{ist}{r} = map_pie(map_ir);
100 self.sn.phases = phases;
101 self.sn.phasessz = max(self.sn.phases,ones(size(self.sn.phases)));
102 self.sn.phasessz(self.sn.nodeToStation(self.sn.nodetype == NodeType.Join),:)=phases(self.sn.nodeToStation(self.sn.nodetype == NodeType.Join),:);
103 % Marked (
MMAP) source classes share
the carrier
's modulating chain: the
104 % non-carrier classes (mark index > 1) contribute a single always-zero
105 % state column rather than their own phase block.
106 if isfield(self.sn,'markidx
') && ~isempty(self.sn.markidx)
107 self.sn.phasessz(self.sn.markidx > 1) = 1;
109 self.sn.phaseshift = [zeros(size(phases,1),1),cumsum(self.sn.phasessz,2)];
113function result = isMAP(proc)
114% ISMAP Check if a process representation is a valid MAP {D0, D1}
116% A valid representation has at least 2 cell elements (D0, D1, plus
117% optional marked/batch matrices D_k as in BMAP/MarkedMAP), all square
118% matrices of the same size.
120if ~iscell(proc) || length(proc) < 2
123n0 = size(proc{1}, 1);
124for e = 1:length(proc)
126 if ~isnumeric(De) || ~ismatrix(De) || size(De,1) ~= size(De,2) || size(De,1) ~= n0
133function MAP = convertToMAP(station, classIdx, proc)
134% CONVERTTOMAP Convert non-Markovian distribution parameters to MAP
136% For non-Markovian distributions, the process representation contains
137% distribution parameters rather than {D0, D1} matrices. This function
138% converts them to an Erlang approximation.
140% Get the distribution object from the station
141if isa(station, 'Source
')
142 dist = station.input.sourceClasses{classIdx}{end};
144 dist = station.server.serviceProcess{classIdx}{end};
147% Get mean for Erlang approximation
148targetMean = dist.getMean();
150% Determine number of phases based on SCV
151% For Det (SCV=0), use high number of phases; for others, match SCV
153if scv < GlobalConstants.CoarseTol
154 % Deterministic or near-deterministic: use 20 phases
157 % Match SCV: for Erlang, SCV = 1/n, so n = 1/SCV
158 nPhases = max(1, ceil(1/scv));
159 nPhases = min(nPhases, 100); % Cap at 100 phases
162% Create Erlang MAP approximation
163MAP = map_erlang(targetMean, nPhases);