Towards Robust Qualitative Spatio-Temporal Reasoning for Hybrid AI Systems

In this short challenge paper, we argue for the need of having robust Qualitative Spatio-Temporal Reasoning (QSTR), as a means to develop hybrid AI systems involving computations with spatio-temporal information. In short, QSTR is a Symbolic AI framework for representing and reasoning about spatial and temporal information via the use of disjunctive natural relations, e.g., “Task A is scheduled after or during Task C” and robustness entails a notion of resistance to the possible future alterations of a spatio-temporal configuration. So far, robustness for QSTR has been defined in the literature in a very rigid manner, in that a configuration may either resist a perturbation or not. Here, we propose an alternative formulation based on probabilities, that should allow a spatio-temporal configuration to also be pliable and adapt to enforced changes. We close the paper by giving some examples of how robust QSTR can be the backbone of hybrid AI systems, emphasizing on abductive learning and AI planning in particular.