Invalid Settings: Conversion to string from embedded.fi is not possible, Reinforcement Learning Designer

I use Reinforcement Learning Designer, MATLAB R2021b, for DDPG agent training with default actor/critic networks of 25 neurons in hidden layers. Ennvironment is generated from Simulink model, and imported to RL Desogner.

I have achieved good training results for RL agent, but when I try to stop training / or training has to finish due to the number of max episodes, this error appears:

"Invalid Settings: Conversion to string from embedded.fi is not possible"

and results of the training are immediately erased. I am not able to find out the reason of the error. What is the reason and how to fix this?Please, help.

Code for environment is below.

obsInfo = rlNumericSpec([3 1],... % dimension of observation vector
    'LowerLimit',[0 -inf  inf]',... 
    'UpperLimit',[inf inf inf]');

obsInfo.Name = 'observations';
obsInfo.Description = 'position, error, integral error';
numObservations = obsInfo.Dimension(1);

actInfo = rlNumericSpec([1 1]);

actInfo.Name = 'voltage';
numActions = actInfo.Dimension(1); 

%Build the environment interface object.
env = rlSimulinkEnv('MY_MLS2EM_rl6_PID_RL','MY_MLS2EM_rl6_PID_RL/DDPG Agent',... 
    obsInfo,actInfo);

[Alexandra Kolosova]

Here is the list of errors I receive in the command window.

> In Simulink.Simulation.internal.DesktopSimHelper

In Simulink.Simulation.internal/DesktopSimHelper/sim

In Simulink/SimulationInput/sim

In rl.env.SimulinkEnvWithAgent>localInnerSimFcn (line 588)

In rl.env.SimulinkEnvWithAgent>@(in_)localInnerSimFcn(simData,in_,[]) (line 251)

In MultiSim.internal.runSingleSim

In MultiSim.internal/SimulationRunnerSerial/executeImplSingle

In MultiSim.internal/SimulationRunnerSerial/executeImpl

In Simulink/SimulationManager/executeSims

In Simulink/SimulationManagerEngine/executeSims

In rl.env/SimulinkEnvWithAgent/executeSimsWrapper (line 229)

In rl.env/SimulinkEnvWithAgent/simWrapper (line 252)

In rl.env/SimulinkEnvWithAgent/simWithPolicyImpl (line 411)

In rl.env/AbstractEnv/simWithPolicy (line 83)

In rl.task/SeriesTrainTask/runImpl (line 33)

In rl.task/Task/run (line 21)

In rl.task/TaskSpec/internal_run (line 166)

In rl.task/TaskSpec/runDirect (line 170)

In rl.task/TaskSpec/runScalarTask (line 194)

In rl.task/TaskSpec/run (line 69)

In rl.train/SeriesTrainer/run (line 24)

In rl.train/TrainingManager/train (line 423)

In rl.train/TrainingManager/run (line 223)

In rl.agent.AbstractAgent/train (line 77)

In rl.internal.app.tool/TrainingSession/startTraining (line 89)

In rl.internal.app/ReinforcementLearningApp/openTrainingSession (line 367)

In rl.internal.app.tab/TrainTab/trainCB (line 224)

In rl.internal.app.tab.TrainTab>@(~,~)trainCB(obj) (line 175)

In internal/Callback/execute (line 128)

In matlab.ui.internal.toolstrip.base/Action/PeerEventCallback (line 828)

In matlab.ui.internal.toolstrip.base.ActionInterface>@(event,data)PeerEventCallback(this,event,data) (line 49)

In hgfeval (line 62)

In javaaddlistener>cbBridge (line 52)

In javaaddlistener>@(o,e)cbBridge(o,e,response) (line 47)