⚡ electrical engineering

Disruption-aware reserve-constrained dispatching of battery-electric AGVs for brownfield manufacturing intralogistics

This study introduces a disruption-aware reserve-constrained dispatching (RCRD) rule for battery-electric AGVs in brownfield factories that screens assignments for charger reachability and ranks them using multiple operational factors, demonstrating a statistically significant reduction in mean tardiness compared to energy-aware baselines, particularly under conditions of simultaneous demand and route stress.

Original authors: Muhammad Rafay Ikram

Published 2026-08-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Muhammad Rafay Ikram

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a busy factory floor not as a static room of machines, but as a living, breathing city where tiny, battery-powered robots zoom around delivering parts. These robots, called Automated Guided Vehicles (AGVs), are the delivery drivers of the manufacturing world. They are essential because they can move materials flexibly, unlike old conveyor belts that are stuck in one place. However, these robots have a major weakness: they run on batteries. If a robot runs out of juice before it can reach a charging station, the whole production line grinds to a halt. Furthermore, factories are messy places. Sometimes a forklift blocks an aisle, a machine breaks down, or a rush order comes in, creating traffic jams and detours. The big question for factory managers is: how do you tell these robots which job to do next? Do you just send the closest one? Do you send the one with the most battery? Or do you send the one that can handle a sudden roadblock without getting stuck? This is the puzzle of "dispatching"—making split-second decisions to keep the factory running smoothly without wasting energy or time.

This paper introduces a clever new rule for making those decisions, called "Disruption-aware Reserve-Constrained Dispatching" (or RCRD for short). Think of it as a super-smart traffic controller for a fleet of electric delivery bots. The researchers wanted to know if giving these robots real-time information about roadblocks and detours actually helps, or if it's just extra noise. To test this, they built a digital twin of a factory with ten workstations, eight robots, and two charging stations. They simulated thousands of scenarios where jobs arrived randomly, and sometimes aisles got blocked.

The new rule works like a careful parent sending a child on an errand. Before sending a robot to pick up a part, it first checks a hard safety rule: "Do you have enough battery to finish the job and still make it to a charger afterward?" If the answer is no, the robot is skipped. If the answer is yes, the rule then ranks the remaining robots based on a score. It considers how far the robot has to drive empty, how much longer the trip will be because of a detour (the "route inflation"), how much battery "cushion" is left, and how late the delivery might be.

The results from the simulation were interesting. The new rule didn't just guess; it actually improved performance. Compared to a standard energy-aware rule that ignored roadblocks, the new RCRD rule reduced the average delay (tardiness) by 6.9%. In the worst-case scenarios—where there was a huge rush of orders and multiple blocked aisles—the improvement jumped to 11.2%. The rule was also incredibly fast, making a decision in just 58.8 microseconds, which is faster than a human blink.

However, the paper is careful not to call this a magic bullet. The improvement wasn't huge in every situation. When the factory was calm or when there were only minor issues, the new rule didn't do much better than the old ones. The real value appeared only when things got chaotic: when high demand met with traffic jams and limited charging spots. The researchers found that the "detour information" was most valuable when the robots had very little room to maneuver (low "slack"). If a robot has plenty of time and battery, knowing about a detour doesn't change much. But when the system is stressed, that extra bit of information helps the robots avoid getting stuck in a dead end or running out of power.

In short, the paper suggests that for factory managers, adding "traffic awareness" to their robot controllers is a smart move, but only if they are dealing with a busy, unpredictable environment. It's not about replacing the whole system with a complex computer program; it's about adding a simple, transparent check that says, "Hey, that path is blocked, and you're low on battery, so maybe don't go there." The study proves that this simple check can save time and keep the production line moving, especially when the factory floor is in a bit of a frenzy.

Technical Summary: Disruption-Aware Reserve-Constrained Dispatching of Battery-Electric AGVs

Problem Statement
In brownfield manufacturing environments, battery-electric Automated Guided Vehicles (AGVs) must navigate dynamic constraints including temporary route restrictions (e.g., maintenance, forklift traffic), finite charging resources, and stochastic task arrivals. A critical operational challenge is ensuring that an AGV assigned to a time-critical task retains sufficient energy not only to complete the transport but also to reach a charger afterward. Existing dispatching methods often suffer from "myopia," where decisions based solely on proximity, due dates, or simple energy thresholds fail to account for the compounding effects of route detours and limited operational slack. The specific research question addressed is: What incremental value does current route-disruption information provide to a dispatching rule that already accounts for energy reserves, urgency, empty travel, and task priority?

Methodology
The study proposes Disruption-aware Reserve-Constrained Dispatching (RCRD), an event-driven, rule-based controller designed for the fleet-management layer. The methodology is characterized by the following components:

Hard Feasibility Screening: Before ranking, candidate vehicle-job pairs are screened against a hard constraint. A pair is feasible only if the vehicle's state of charge (SoC) after the task allows it to reach the nearest available charger with a safety reserve ( $s_{safe}$ ) of 8 SoC percentage points.
Normalized Scoring Function: Feasible candidates are ranked using a weighted score ( $C_{vj}$ $C_{v j}$ ) that normalizes five attributes:
1. Empty travel distance ( $E_{vj}$ ).
2. Route inflation ( $I_{vj}$ ): The excess distance caused by detours compared to the nominal path.
3. Soft reserve headroom ( $R_{vj}$ ): A penalty for pairs with minimal margin above the hard safety requirement.
4. Projected tardiness ( $T_{vj}$ ).
5. Task priority ( $p_j$ ).
  The coefficients for these terms were locked prior to confirmatory testing to prevent overfitting.
Experimental Design: The study employs a matched ablation design using Discrete-Event Simulation (DES). A "Matched Energy-Aware Core" comparator uses the exact same logic, coefficients, and candidate sets as RCRD but sets the route-inflation coefficient to zero. This isolates the specific value of route-state information.
Validation: The evaluation utilizes 240 matched seed-profile observations across four operating profiles (Nominal, Demand Surge, Aisle Disruption, and Combined Stress) and varying resource capacities (AGV/charger counts). Statistical significance is assessed via paired bootstrap confidence intervals and Wilcoxon signed-rank tests.

Key Results

Overall Performance: RCRD achieved the lowest mean tardiness (45.64 time units) among four tested policies, including FIFO-nearest, EDD-nearest, and the Matched Energy-Aware Core. Compared to the Matched Core, RCRD reduced mean tardiness by 6.9% (3.37 time units).
Statistical Significance: The paired reduction in mean tardiness had a 95% bootstrap confidence interval of [1.00, 5.78], which excludes zero. However, improvements in 95th-percentile tardiness, on-time rates, and empty-travel share were not statistically significant across all scenarios.
Scenario Dependence: The value of route-information is highly context-dependent.
- Combined Stress: Under simultaneous high demand and aisle disruptions, RCRD reduced mean tardiness by 11.2%.
- Resource Bottlenecks: The advantage increased significantly under a one-charger bottleneck configuration, where RCRD improved mean tardiness by 14.00 time units compared to the core.
- Low Stress: In nominal or isolated disruption scenarios, the performance difference was negligible or slightly negative, indicating that route information is not universally beneficial.
Computational Efficiency: The mean decision latency for RCRD was 58.8 microseconds, with a 95th percentile of 81.6 microseconds, confirming the rule's suitability for real-time fleet control without requiring iterative optimization.

Significance and Claims
The paper positions RCRD not as a global optimizer or a learning-based policy, but as an implementable, transparent decision rule for manufacturing engineers and fleet managers. Its primary contribution is the isolation of the "information value" of route disruptions.

The author claims that:

Contextual Value: Route-state information is most valuable when disturbances interact with limited operational slack (e.g., high demand or charging bottlenecks). In stable environments, the added complexity yields little benefit.
Practicality: The rule can be deployed in existing Manufacturing Execution Systems (MES) or Warehouse Control Systems (WCS) without retraining or heavy computational overhead, provided the system can supply detour estimates.
Conditional Improvement: The study modestly concludes that RCRD offers a "conditional engineering evidence" for delay reduction rather than universal policy dominance. It serves as a low-cost control feature that becomes critical when detours and resource contention coincide.

The paper explicitly states that deployment requires plant-specific calibration of coefficients and safety validation, and it does not claim to replace incumbent policies in all scenarios.

Technical Summary: Disruption-Aware Reserve-Constrained Dispatching of Battery-Electric AGVs

More like this