Conference Agenda

Nowadays, a diverse range of business sectors necessitates the development of optimal planning that aligns with various business-oriented objectives and constraints Consequently, generating scheduling and routing solutions for organizational staff has become exceedingly complex, especially with large-scale realistic problems. Optimization models play a vital role in assisting organizational decision-makers for addressing operational challenges and generating satisfactory solutions using powerful optimizers. This research work tackles the challenges associated with realistic Scheduling and Routing Problems (SRPs) with a multi-day horizon. To address these challenges, BL.Optim is introduced, which is an OptaPlanner-based optimizer with a model-based architecture. BL.Optim aims to capture routing and scheduling requirements for a wide category of Constraint Satisfaction Problems (CSP), with customizable requirements that can activate necessary constraints according to specific realistic use cases. A set of pre-formulated CSP-based soft constraints and non-violated hard constraints are embedded within BL.Optim. To assess the optimality of solutions generated by BL.Optim, two real-life use cases of Preventive Maintenance SRPs in Spain and France are presented, involving up to 481 activities to be performed by 18 staff members over 4 weeks. These use cases are tested in BL.Optim under different metaheuristic configurations (e.g., tabu search), and the solution scores are compared. Experimental results demonstrate high-quality optimized solutions free from hard constraint violations and with minimal penalties for soft constraints, achieved within a short computing time of just 8 minutes.

We present the architecture and decisions behind the OR framework for energy trading operations in a power-sector company.

With established processes and software ecosystem, there are many path-dependent constraints but also room for some fundamental choices.

We discuss the reasons behind the architectural decisions taken.

In particular, we highlight our experience using a single strongly statically typed functional programming language for all aspects from modelling and algorithm development to deployment and interactive data analysis.

Further points are API and library integration including dependency management (both up- and downstream), documentation vs specification and the ubiquitous datastore for algorithmic IO data.