I developed a configurable fleet simulation platform for Robust AI that creates a detailed virtual recreation of a customer’s future Carter operation. A complete model can be built in less than 30 minutes, allowing the Solutions team to run dozens of fleet, labor, and workflow scenarios every hour.

The simulator combines facility layouts, operational data, and a custom Carter behavior library to evaluate how a proposed solution will perform before deployment.

Challenge
AMR solutions are often sized using spreadsheet-based calculations that rely on averages for travel, productivity, and robot utilization. These tools can provide an initial estimate, but they cannot fully capture the interactions between Carters, associates, inventory, congestion, induction, unloading, and changing order demand.

Robust AI needed a higher-fidelity approach that could validate fleet requirements while clearly illustrating key operational concepts to customers. At the same time, the tool needed to be extremely fast. Proposals are often developed within days, leaving little time to construct a traditional simulation model or manually test every potential design.

Approach
I built a reusable simulation framework that automatically generates warehouse racking and storage locations from AutoCAD layouts, significantly reducing the time required to construct each model.

A custom object library recreates Carter’s navigation, mission logic, picking workflow, and interactions with warehouse associates. The simulator is then configured using findings from customer data analysis, including:
- Hourly order volume
- Order and SKU profiles
- Units and lines per order
- Inventory velocity and slotting
- Pick, induct, and unload time studies
- Labor schedules and availability
- Operational rules and process constraints

Together, these inputs create a realistic representation of the proposed future-state operation rather than relying on generalized spreadsheet assumptions.

The framework was designed for rapid iteration. Once configured, users can quickly test changes to fleet size, staffing, slotting, workflow logic, and operating assumptions. Built-in dashboards track throughput, cycle times, travel, queueing, utilization, and mission performance across the entire system.

The visual model also helps customers understand concepts that are difficult to communicate through a spreadsheet, such as robot-associate interactions, congestion, bottleneck formation, and the operational impact of different fleet or staffing decisions.

Outcome
The platform reduced model development to less than 30 minutes and made it possible to evaluate dozens of scenarios within a single hour.

This provides the fidelity of a detailed simulation without sacrificing the speed required to support proposals developed on compressed timelines. The Solutions team can identify bottlenecks earlier, compare alternatives more quickly, and support fleet recommendations with a data-backed recreation of how Carter is expected to perform.

The simulator also creates a more compelling customer experience by turning abstract assumptions and calculations into a clear visual demonstration of the proposed future-state operation.