Modelling Reactive and Proactive Behaviour in Simulation

Mazlina Abdul Majid, Peer-Olaf Siebers, Uwe AickelinarXiv 2010 · OpenReview

TL;DR

This paper explores how to best simulate human behavior in complex systems, specifically comparing traditional Discrete Event Simulation (DES) with a combined DES and Agent-Based Simulation (ABS) approach. The study found that while both methods can model basic proactive behaviors, the combined DES/ABS approach offers a more nuanced representation, particularly in capturing the impact of proactive staff actions on system performance.

What Problem Does This Paper Address?

Simulating human behavior is crucial for understanding and optimizing complex systems, especially in service industries like retail. However, existing simulation methods have limitations in accurately capturing the full spectrum of human actions. Specifically, traditional Discrete Event Simulation (DES) struggles to model "proactive" behavior – actions taken by individuals without direct prompting, driven by initiative. Agent-Based Simulation (ABS) is better suited for this, but integrating it with DES, which excels at modeling system-level processes, presents a challenge. This paper aims to bridge this gap by investigating how a combined DES/ABS approach models both reactive (prompted) and proactive human behaviors in a retail fitting room scenario, and how these models compare to a pure DES approach.

Methods

The researchers developed two simulation models for a fitting room in a women's wear department:

  • Discrete Event Simulation (DES) Model: This model treated customers, staff, and fitting rooms as passive objects, focusing on the sequential flow of events.
  • Combined Discrete Event and Agent-Based Simulation (DES/ABS) Model: This approach used DES for the overall system process (queues, fitting room availability) but modeled customers, staff, and fitting rooms as active "agents." This allowed for more autonomous and interactive behavior.

Both models were implemented in the AnyLogic™ simulation software. The core of the study involved:

  1. Modeling Reactive Behavior: Staff counting garments, providing help, and collecting items.
  2. Modeling Proactive Behavior: Staff speeding up service by 20% when the fitting room became busy or when customers were waiting. This was implemented using decision rules and probabilistic distributions.
  3. Sensitivity Analysis: Varying customer arrival rates to observe the impact on performance measures (waiting times, staff utilization, etc.) for both models.
  4. Statistical Comparison: Comparing the results of a previous experiment (Experiment A, reactive behavior only) with the current experiment (Experiment B, mixed reactive and proactive behavior) using the Mann-Whitney statistical test to assess differences in customer waiting times and staff utilization.

Datasets

The study used observational data from a real-world departmental store's fitting room to model customer arrival rates. This data was used to create an exponential distribution with an hourly changing arrival rate for the simulation models. No new benchmarks were introduced; the focus was on comparing existing simulation methodologies.

Key Insights and Ideas

  • Proactive behavior matters: Even simple proactive actions by staff (like speeding up service) can have a noticeable impact on system performance, influencing metrics like staff utilization.
  • DES/ABS offers richer representation: The combined DES/ABS model provided a more detailed and potentially more accurate representation of human behavior, especially proactive actions, compared to a traditional DES model.
  • Bottlenecks are critical: The study highlighted that even with proactive staff behavior, system bottlenecks (like the number of fitting room cubicles) can limit the overall improvement in customer waiting times.
  • Similarities in core behavior: For basic reactive behaviors, both DES and DES/ABS models showed similar output performance, suggesting that for simpler scenarios, traditional DES might suffice.

Key Results

  • Sensitivity Analysis: Both DES and DES/ABS models showed increased performance measures (e.g., waiting times) with higher customer arrival rates. The DES/ABS model consistently served more customers and had higher staff and cubicle utilization compared to the DES model.
  • Statistical Comparison (Reactive vs. Mixed Reactive/Proactive):
    • Customer Waiting Times: The statistical tests indicated no significant difference in average customer waiting times between the reactive-only and mixed reactive/proactive experiments for both DES and DES/ABS models. This was attributed to the fitting room cubicles acting as a bottleneck.
  • Staff Utilization: A significant difference was found in staff utilization. The mixed reactive and proactive behavior experiment (Experiment B) resulted in lower staff utilization compared to the reactive-only experiment (Experiment A) for both DES and DES/ABS models. This suggests that proactive behavior can lead to more efficient staff deployment.

Limitations and Future Work

  • Limitations:
    • The study acknowledges that the impact of proactive behavior on customer waiting times was limited due to the fitting room cubicle bottleneck.
  • The proactive behavior modeled was relatively simple (speeding up service).
  • Future Work:
    • Investigate further differences between DES and DES/ABS models, particularly in sensitivity analysis experiments.
  • Include other forms of proactive staff behavior, such as staff asking for help from colleagues when encountering difficult situations.
  • Apply the combined DES/ABS approach to a public sector case study to assess its generalizability for human-centric complex systems.

Summary updated

Abstract

This research investigated the simulation model behaviour of a traditional and combined discrete event as well as agent based simulation models when modelling human reactive and proactive behaviour in human centric complex systems. A departmental store was chosen as human centric complex case study where the operation system of a fitting room in WomensWear department was investigated. We have looked at ways to determine the efficiency of new management policies for the fitting room operation through simulating the reactive and proactive behaviour of staff towards customers. Once development of the simulation models and their verification had been done, we carried out a validation experiment in the form of a sensitivity analysis. Subsequently, we executed a statistical analysis where the mixed reactive and proactive behaviour experimental results were compared with some reactive experimental results from previously published works. Generally, this case study discovered that simple proactive individual behaviour could be modelled in both simulation models. In addition, we found the traditional discrete event model performed similar in the simulation model output compared to the combined discrete event and agent based simulation when modelling similar human behaviour.
BibTeX
@misc{majid2010modelling,
  author = {Mazlina Abdul Majid and Peer-Olaf Siebers and Uwe Aickelin},
  title = {{Modelling Reactive and Proactive Behaviour in Simulation}},
  year = {2010},
  eprint = {1006.3652},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1006.3652},
}
Figure