- fun_to_cons_var(+CostFun, -ConsList, -CostVar)
- Convert a cost expression to a variable and a list of constraints, suitable to pass into probe_cstr_sched/7
- probe_cstr_sched(+Starts, +Durations, +Resources, ++MaxResource, +Constraints, -Cost, ++Options)
- Find a resource-feasible schedule that minimises the cost, subject to the constraints
- probe_sched(+Starts, +Durations, +Resources, ++MaxResource, ?CostFun)
- Find a resource-feasible schedule that minimises the cost function
The user interface is similar to the cumulative constraint, with a list of task start times, durations and resources; and a maximum resource limit. There is one extra argument, a cost function which is to be minimised.
Probing for scheduling differs radically from the cumulative constraint because it includes a search routine. In its behaviour it is an optimisation procedure and not simply a constraint which enforces consistency.
The search is focussed towards an optimal vlaue of the cost function.
% Example program: simple schedule optimisation :- lib(probing_for_scheduling). :- lib(fd). ex1([X,Y,Z],Cost) :- [OldX,OldY,OldZ]=[1,5,5], Durations=[10,5,5], Resources=[1,2,1], MaxResource=2, NewStarts=[X,Y,Z], fd:(NewStarts::1..10), CostFun= abs(X-OldX) + abs(Y-OldY) + abs(Z-OldZ), probe_sched(NewStarts,Durations,Resources,MaxResource,CostFun), Cost is CostFun. ex2([X,Y,Z],Cost) :- [OldX,OldY,OldZ]=[1,5,5], Durations=[10,5,5], Resources=[1,2,1], MaxResource=2, NewStarts=[X,Y,Z], fd:(NewStarts::1..10), Constraints=[ XDiff >= X-OldX, XDiff >= OldX-X, YDiff >= Y-OldY, YDiff >= OldY-Y, ZDiff >= Z-OldZ, ZDiff >= OldZ-Z, Cost =:= XDiff+YDiff+ZDiff ], Options=[granularity(1),priority(5)], probe_cstr_sched(NewStarts,Durations,Resources,MaxResource, Constraints,Cost,Options). ex3(Starts,MaxResource,Cost) :- Starts=[S1,S2,S3,S4], fd:(S1::1..10), fd:(S2::3..10), fd:(S3::1..5), fd:(S4::3..10), [D1,D2,D3,D4]=[5,5,3,3], Durations=[D1,D2,D3,D4], Resources=[1,1,1,1], CostFun= max([S1+D1,S2+D2,S3+D3,S4+D4]), probe_sched(Starts,Durations,Resources,MaxResource,CostFun), Cost is CostFun.