Dispatching Policies for Levitated Shuttle Fleets in a Robotic Production Setup
Kajuutti, Eevi (2026)
2026
Robotiikan kandidaattiohjelma - Bachelor’s Programme in Robotics
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
Hyväksymispäivämäärä
2026-05-12
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202605115365
https://urn.fi/URN:NBN:fi:tuni-202605115365
Tiivistelmä
This Bachelor’s thesis evaluates dispatching policies for levitated shuttle fleets in a robotic production setup. The study addresses three research questions: (RQ1) which dispatching policies, KPIs, and test scenarios are suitable for studying shuttle-to-robot task dispatching; (RQ2) how dispatching policies affect system performance under different fleet-sizes and workloads; and (RQ3) what practical design guidelines can be derived for real production deployment, including key limitations and improvement directions.
A cycle-based simulation was implemented in Automation Studio 6, where the control logic was written in Structured Text (ST) and executed in a PLC-style cyclic task to compare a baseline First-Come-First-Served policy with fixed round-robin workstation assignment (RR) against two alternative policies: FCFS with workstation availability and queue awareness (AVAI) and Nearest- Vehicle-First (NVF). In addition, a system-variant scenario was evaluated to assess the impact of modified task allocation constraints. The experiments covered 36 scenarios (three workloads, three fleet-sizes, and four policy/variant cases), with five independent replications per scenario to account for variability. Performance was evaluated using productivity, delay, and congestion KPIs, including throughput, average and P95 completion time, robot idle time, workstation queue length, and a congestion indicator.
The results show that increasing the number of shuttles does not necessarily improve performance: the system is congestion-limited, and excessive fleet-size increases blocking in shared resources (e.g., pick-up and drop-off segments). An optimal fleet-size emerged in the tested range, and the best configuration depended on both workload and dispatching policy. The RR baseline achieved the highest peak throughput in this simplified setup, whereas NVF provided the most stable behavior across configurations, offering a practical trade-off between peak performance and predictability. The system variant significantly degraded performance, highlighting the sensitivity of congestion-limited systems to structural constraints.
A cycle-based simulation was implemented in Automation Studio 6, where the control logic was written in Structured Text (ST) and executed in a PLC-style cyclic task to compare a baseline First-Come-First-Served policy with fixed round-robin workstation assignment (RR) against two alternative policies: FCFS with workstation availability and queue awareness (AVAI) and Nearest- Vehicle-First (NVF). In addition, a system-variant scenario was evaluated to assess the impact of modified task allocation constraints. The experiments covered 36 scenarios (three workloads, three fleet-sizes, and four policy/variant cases), with five independent replications per scenario to account for variability. Performance was evaluated using productivity, delay, and congestion KPIs, including throughput, average and P95 completion time, robot idle time, workstation queue length, and a congestion indicator.
The results show that increasing the number of shuttles does not necessarily improve performance: the system is congestion-limited, and excessive fleet-size increases blocking in shared resources (e.g., pick-up and drop-off segments). An optimal fleet-size emerged in the tested range, and the best configuration depended on both workload and dispatching policy. The RR baseline achieved the highest peak throughput in this simplified setup, whereas NVF provided the most stable behavior across configurations, offering a practical trade-off between peak performance and predictability. The system variant significantly degraded performance, highlighting the sensitivity of congestion-limited systems to structural constraints.
Kokoelmat
- Kandidaatintutkielmat [11997]