In forward chaining, a model controller makes a prediction from the RHS and also makes an imdl to show how the model was instantiated. The example below is for the reuse model M6 in hand-grab-sphere (simplified):
fact_281 matches the LHS and binds variables T0 and T1 to the timings 200ms and 800ms. Note that variables T0 and T1 are also in the model's template arguments. Then the controller matches with requirement fact128 where the template arguments have the more "narrow" time interval 600ms to 700ms. This updates variables T0 and T1 to these more narrow values. Then the forward guards are used to compute the RHS time interval variables T2 and T3 which you can see in prediction fact_290 as 700ms and 800ms. So far, so good. Then the controller makes imdl fact_291 to show how the model was instantiated to make the prediction. But the template arguments are still 200ms and 800ms. These are not the values that were used to make the prediction. They should be 600ms and 700ms.
When the model controller makes the imdl fact_291 the binding map has the correct values for T0 and T1, but these haven't been used to update the imdl.
jefft0
commented
1 year ago
In forward chaining, a model controller makes a prediction from the RHS and also makes an imdl to show how the model was instantiated. The example below is for the reuse model M6 in hand-grab-sphere (simplified):
fact_281matches the LHS and binds variablesT0andT1to the timings 200ms and 800ms. Note that variablesT0andT1are also in the model's template arguments. Then the controller matches with requirementfact128where the template arguments have the more "narrow" time interval 600ms to 700ms. This updates variablesT0andT1to these more narrow values. Then the forward guards are used to compute the RHS time interval variablesT2andT3which you can see in predictionfact_290as 700ms and 800ms. So far, so good. Then the controller makes imdlfact_291to show how the model was instantiated to make the prediction. But the template arguments are still 200ms and 800ms. These are not the values that were used to make the prediction. They should be 600ms and 700ms.When the model controller makes the imdl
fact_291the binding map has the correct values forT0andT1, but these haven't been used to update the imdl.