Automated planning systems are increasingly deployed in real-world applications, often characterised by uncertainty and noise stemming from sensors, actuators, and environmental conditions. Under such circumstances, improving the deployability of generated plans requires assessing their robustness to varying conditions, thereby reducing the need for costly replanning. Replanning can be computationally intensive and may hinder the practical applicability of planning systems. In many domains, such as urban traffic control or underwater exploration, it is often sufficient for plans to reach an acceptable region rather than the exact goal. A key distinction in this context lies between valid plans (which achieve the intended goal under ideal conditions) and executable plans (which remain feasible under uncertainty or perturbation). This paper formalises the notion of execution-invariant planning tasks, in which plans are robust to noise and uncertainty. To foster the adoption of automated planning in real-world settings, we propose a statistical framework for evaluating plan robustness, offering a quantifiable measure of a plan’s ability to reach a goal within a specified tolerance under diverse perturbations or uncertainty. We validate our approach in two real-world domains, demonstrating its effectiveness.

The problem of traffic signal optimisation has been successfully tackled using the PDDL+ planning formalism, which also provides an ideal ground for simulating traffic behaviour and performing what-if analysis to assess and compare alternative scenarios. This line of research leads to approaches that can efficiently generate high-quality signal plans with significant benefits in terms of congestion and emissions reduction, as demonstrated both in simulations and real-world deployments. Existing models for automated planning-based traffic signal control can be roughly divided into two classes. (i) Models maximising the flexibility of the traffic controller, where the planning system can dynamically adjust the duration of traffic stages without constraints on the overall cycles and on the differences between subsequent cycles. (ii) Models that guarantee the deployability of traffic signal control techniques also on legacy infrastructure, by forcing the AI approach to select the cycle configurations of traffic signals for the controlled junctions from a given pre-defined set. Of course, both classes offer valuable properties and benefits: the extreme flexibility helps shed light on the potential gains achievable through investment in brand-new infrastructure, whereas the deployable approaches ensure the immediate usability of tools to maximise short-term impact. To bridge the gap between different model classes and explore the trade-off between flexibility and deployability, we present the Trade model. It enables the enforcement of key constraints required for deployability while preserving a level of flexibility that surpasses the capabilities of traditional traffic control infrastructure.