Dynamic Request Handling in Vehicle Routing and Logistics

52 concepts52 papers1 proposal

Traditional logistics and transit systems operate under a convenient mathematical fiction: the static demand assumption. In this paradigm, every passenger trip, package delivery, or service request is assumed to be known before a vehicle ever leaves the depot. For instance, classic formulations of the Periodic Vehicle Routing Problem (Jianfeng Zhu et al. in A Vehicle Routing System to Solve a Periodic Vehicle Routing Problem for a Food Chain in Hong Kong) or strategic transit design (David Bergman et al. in The Integrated Last-Mile Transportation Problem (ILMTP)) rely heavily on pre-calculated historical data or upfront snapshots.

While this assumption makes the underlying mixed-integer linear programs (MILPs) tractable, it creates brittle systems. If a traffic accident occurs, a passenger books a ride on the fly, or a new delivery contract arrives mid-cycle (Jia Zhang et al. in Efficient Delivery Policy to Minimize User Traffic Consumption in Guaranteed Advertising), a static schedule cannot adapt. To build resilient, real-world systems, researchers are shifting toward dynamic request handling—algorithms that treat routing as a continuous, online optimization problem.

Why Dynamic Adaptation Works: The Core Intuition

If we do not know the future, how can we optimize for it? The core insight behind dynamic request handling is that we do not need perfect foresight to achieve near-optimal efficiency. Instead, we can exploit spatial-temporal locality and flexibility margins.

Consider a vehicle already in transit. When a new request arrives, a naive approach would re-solve the entire routing problem from scratch, which is computationally impossible in real-time. Dynamic systems instead use reactive scheduling. Remarkably, Bryan David Galarza Montenegro et al. demonstrated in The real-time dynamic online feeder service with a maximum headway at mandatory stops that real-time optimization frameworks can achieve performance levels within 6.5% of an omniscient static model.

This works because urban transit demand is rarely completely random; it possesses structural properties that can be managed via clever algorithmic abstractions rather than global re-optimization.

Mechanisms of Real-Time Adaptation

To break free from static planning, modern frameworks employ three primary mechanisms to handle streaming, real-time requests:

1. Batching and Rolling Horizons

Instead of treating requests as a single static block or processing them purely one-by-one (which leads to myopic decisions), streaming requests are grouped into tight, sequential time windows.

Alexandre M. Florio et al. show in A Hierarchical Grouping Algorithm for the Multi-Vehicle Dial-a-Ride Problem that partitioning continuous, streaming request arrivals into hierarchical batches allows algorithms to maintain global efficiency while providing worst-case performance guarantees. Similarly, Xiaohui Bei et al. in Algorithms for Trip-Vehicle Assignment in Ride-Sharing highlight the necessity of moving away from static snapshots to handle the continuous, stochastic nature of passenger demand.

2. Semi-Flexible Routing and Slack Management

Rather than choosing between rigid fixed routes and chaotic fully-flexible routing, systems can embed "slack" into their schedules to absorb real-time demand fluctuations.

Dilay Aktaş et al. in A variable neighborhood search algorithm for a public bus line with a demand-responsive operation during peak hours design bus lines that use pre-planned shortcuts during low-demand segments to free up capacity for real-time passenger surges. By guaranteeing service at mandatory stops while dynamically routing to optional stops based on live bookings (Bryan David Galarza Montenegro et al. in A demand-responsive feeder service with a maximum headway at mandatory stops), the system adapts on the fly without violating maximum wait-time constraints.

3. Hybrid Learning and Optimization

To handle the massive state space of real-time routing, researchers combine the speed of machine learning with the precision of classical optimization.

Jiayu Chen in DeepFreight: A Model-free Deep-reinforcement-learning-based Algorithm for Multi-transfer Freight Delivery proposes a hybrid framework where multi-agent reinforcement learning handles immediate, dynamic request arrivals, while a Mixed-Integer Linear Programming (MILP) "cleanup" phase ensures operational constraints are strictly met. This allows the system to scale to real-time environments where pure optimization solvers would time out.

From On-Demand Transit to Infrastructure

Moving past the static demand assumption changes more than just vehicle paths; it fundamentally alters how we size and deploy infrastructure. As Ramón Auad notes in Ridesharing and Fleet Sizing For On-Demand Multimodal Transit Systems, transitioning from static models to online algorithms allows operators to dynamically size fleets, making multimodal transit systems both cheaper and faster.

Whether managing the unpredictable arrival of electric vehicles at charging stations (Nathaniel Tucker et al. in An Online Pricing Mechanism for Electric Vehicle Parking Assignment and Charge Scheduling) or rebalancing bike-sharing stations during peak hours (Yiwei Fan et al. in Distributed forecasting and ant colony optimization for the bike-sharing rebalancing problem with unserved demands), breaking the static assumption is the key to unlocking true urban mobility.

Key Foundational Work

Real-Time Multivehicle Truckload Pickup and Delivery Problems (2004) — contributed 2 concept elements to this direction.
A Vehicle Routing System to Solve a Periodic Vehicle Routing Problem for a Food Chain in Hong Kong (2008) — contributed 1 concept elements to this direction.
A combinatorial allocation mechanism with penalties for banner advertising (2008) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_static_demand_assumption_2dd7338e,
  title = {static demand assumption},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/static-demand-assumption/rdr_2dd7338ef1e845ebb048fa45f6b065ab},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_static_demand_assumption_2dd7338e_v1,
  title = {static demand assumption},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/static-demand-assumption/rdr_2dd7338ef1e845ebb048fa45f6b065ab/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake