Configuration of Abstractions. The Year 1 deliverable [BM15] describes three abstractions (also referred to as model approximations) for model checking of co-models consisting of both, DE and CT systems. Here, we briefly repeat the details of these abstractions and their current status in the RTT-MBT model checker 2. . Interval Abstraction uses the well-known box or interval domain [CC77] to bound the range of continuous variables. RTT-MBT allows to specify upper and lower bounds for all variables of a co-model, and both the model checker and the test case generator allow input variables to hold values only in the provided range. An example of an interval abstraction for a continuous variable is given in Fig. 1. To implement this functionality for a variable v, the component that computes abstracted value has to be removed from the input to RTT-MBT. Suppose that lv and uv, respectively, denote the lower and upper bounds of v. The transition relation is conjoined with an additional constraint: k lv ≤ vi ≤ uv ∈ V C
Configuration of Abstractions. The Year 1 deliverable [BM15] describes three abstractions (also referred to as model approximations) for model checking of co-models consisting of both, DE and CT systems. Here, we briefly repeat the details of these abstractions and their current status in the RTT-MBT model checker 2. ^ Interval Abstraction uses the well-known box or interval domain [CC77] to bound the range of continuous variables. RTT-MBT allows to specify upper and lower bounds for all variables of a co-model, and both the model checker and the test case generator allow input variables to hold values only in the provided range. An example of an interval abstraction for a continuous variable is given in Fig. 1. To implement this functionality for a variable v, the component that computes abstracted value has to be removed from the input to RTT-MBT. Suppose that lv and uv, respectively, denote the lower and upper bounds of v. The transition relation is conjoined with an additional constraint: k lv ≤ vi ≤ uv ∈ V C Let V denote the set of variables to be abstracted, and assume that each v is computed by a component v. Further, let remove{C1,...,Cn} denote the operation that removes a set of components from the transition relation. Then, the overall formula to be solved is given by: k−1 k I ∧ ^ remove{Cv|v∈V}(T (si, si+1)) ∧ ¬ϕ ∧ ^ ^(lv ≤ vi ≤ uv) i=0 v∈V Intuitively, the computation of a variable v is removed from the transition relation, and in each unrolling step a value that satisfies the bounds constraint may be chosen. Gradient-Based Interval Abstraction is more constrained than interval abstraction as it too specifies to what extent the value of an input variable may change in a given time unit, which has to be specified manually by the designer of the abstraction. For instance, interval abstraction allows a variable to first equal its lower bound, and then, in the next computation 2The current release of RTT-MBT provides stable support for interval-based abstrac- tion.
