# Real-World, Large Scale, Multi-Period Log Truck Routing and Scheduling : Application to Canadian Forestry

arXiv: https://arxiv.org/abs/2607.16978 · HTML: https://mlsanity.com/paper/2607.16978/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 18 Jul 2026.

## Built on by (most cited first, 1 of 1)

- [Learning to recover: Adaptive local branching with reinforcement learning for log-truck routing and scheduling under disruptions](https://mlsanity.com/paper/2607.20901/) (2026)

## Abstract (arXiv)

This paper addresses the multi-period log-truck routing and scheduling problem (LTRSP), a key operational activity in the forestry industry, where transportation accounts for more than one-third of total operational costs. The Canadian forestry sector faces significant logistical difficulties driven by vast geographic distances, seasonal variability, volatile markets, and environmental considerations. Our research tackles a long-standing open question in the forestry operations literature, namely the absence of an exact and scalable formulation of the forestry vehicle routing problem integrating the full range of operational constraints citeronnqvist2015operations: textitHow can we model and solve an exact formulation of the forestry VRP problem? In response, we analyze business rules specific to the forestry sector to construct a routing network reflecting real operational practices, and then propose a comprehensive improved mixed-integer linear programming (MILP) formulation incorporating all known forestry operational constraints, with a detailed justification of key modeling choices such as time discretization, along with a decomposition-based solution methodology tailored for large-scale, multi-period industrial instances. To address the inherent combinatorial complexity, we combine state-of-the-art solvers with a metaheuristic decomposition strategy based on textitRelax&Fix and textitFix&Optimize. Computational experiments on historical data from a Canadian forest company demonstrate near-optimal results within practical computation times, with significant financial gains and reduced greenhouse gas emissions.