A simulated annealing algorithm is now an option in place of a genetic algorithm for agent decision-making. A new vignette and new documentation explains how and when to use simulated annealing in GMSE
A new argument permits adjustment to the mutation magnitude in the genetic algorithm and simulated annealing algorithm.
A proportion of landscape yield can now be added to user and
manager budgets with the new arguments
The manager action threshold (which determines whether
intervention is warranted based on the observed population condition)
can now be based on population trajectory rather than current observed
density. New arguments include
An improved GUI with
Error catches have been added to check the structure of inputs in
Rare cases of dead resources feeding in the subsequent time step have been fixed.
The order of operations in the resource submodel has changed. Formerly, resources would move on the landscape, reproduce, die, and then potentially consume on the landscape. Now, resources feed on the landscape, reproduce, die, and then move.
The default landscape ownership is now divided using a shortest-splitline algorithm, which generates more realistic rectangular blocks of land.
The default perceived effects of different actions have been changed for both the manager and users in the fitness function of the genetic algorithm.
Resource birth and death can now be caused by resource
consumption on the landscape using the new arguments
times_feeding (note that movement to new cells occurs
between times feeding).
Managers can apply an action threshold, which stops them from
updating policy in a time step if the observed resource density is
sufficiently close to the target. If they do not update policy, managers
can receive a budget bonus that can be applied to the next time step.
Relevant arguments are
budget_bonus, and a new vignette explains their
The perceived effects of different actions for users can now be
manually set using the arguments,
perceive_kill. Hence, it is
possible to make users perceive some actions to be more or less
effective than they are likely to be.
Land ownership can be allocated unevenly among users using the
Manager and user budgets can now be incremented as a function of
landscape yield using the
gmse_gui function has been updated so that
options available in
Differences between default arguments of
gmse_apply have now been rectified. Now a run of
gmse and looped
gmse_apply with default
subfunctions returns the same system dynamics.
A crash in
gmse_apply caused by setting an argument
name to the argument itself (e.g.,
stakeholders = stakeholders) was noticed sometimes. An
error message is now produced in
gmse_apply to warn against
having the same name for an argument and the argument’s value.