Give us your feedback
TUPLES
TrUstworthy Planning and scheduling with Learning and ExplanationS
Website
Industry Sector
Manufacturing . Electricity, gas, steam and air conditioning supply . Public Administration
Project Timeline
October 1, 2022 – September 30, 2025
The project will take a more integrated and human-centered approach to the development of P&S tools, in order to increase confidence in these systems and accelerate their adoption. In particular, we will design methods that combine the power of data-driven and knowledge-based symbolic AI.
TUPLES intends to showcase these advancements in Trustworthy AI through practical use-cases, including manufacturing, aircraft operations, sport management, waste collection, and energy management, fostering confidence and accelerating adoption in diverse domains.
The cornerstones of our scientific contributions in Trustworthy AI will be:
-combining symbolic P&S methods with data-driven methods to benefit from the scalability and modelling power of the latter, while gaining the transparency, robustness, and safety of the former;
-developing rigorous explanations and verification approaches for ensuring the transparency, robustness, and safety of a sequence of interacting machine learned decisions. Both of these challenges are at the forefront of AI research.
We will demonstrate and evaluate our novel and rigorous methods in a laboratory environment, on a range of use-cases in manufacturing, aircraft operations, sport management, waste collection, and energy management.
EXPECTED IMPACT
OUTCOME 1
To develop hybrid planning and scheduling methods that combine the efficiency, flexibility, and adaptability of data-driven learning approaches with the robustness, reliability, and clarity of model-based reasoning methods. This will require the ability to integrate learned models into the core of current planning and scheduling approaches that rely on constraint satisfaction, combinatorial optimization, and heuristic search algorithms.
OUTCOME 2
To develop verification and explanation methods capable of reasoning about the properties of the solutions produced by planning and scheduling systems, in particular when these are represented by neural networks.
