Trolley And Me | Devpost

Modelo de la banda
GIF
Banda en funcionamiento
Aplicación de entrenamiento

Inspiration

We were inspired by the thought of employees preparing thousands of meals and millions of products every day, under incredible time pressure, where a few seconds and a single subjective decision can cost thousands in waste.

What it does and How we built it

Our solution tackles two critical challenges in airline catering logistics: inconsistent alcohol bottle handling and manual quality control during packing. We deliver a two-part system designed for instant decision-making, real-time error detection, and maximum employee efficiency.

Vision-Guided Alcohol Management System This component transforms the subjective process of evaluating returned alcohol bottles into an automated, data-driven decision. It ensures consistency, reduces high-value waste, and guarantees compliance with airline Service-Level Agreements (SLAs).

Key Deliverables: Physical Prototype (Hardware): A scaled conveyor belt system with a small motor and camera hardware simulating the inspection line. Trained Vision Model (Code/Logic): AI model trained to detect bottle type, estimate fill level, and assess physical condition (e.g., seal integrity, label cleanliness). This model instantly cross-references bottle data against stored airline SLAs to determine the correct action: Keep, Refill, or Discard. Digital Dashboard (App Code): An application visualizing the process, showing real-time metrics on what action to take with the bottle in order to be aligned with the compliance provided by each airline or product category.

AI Voice Assistant for Real-Time Packing Validation This solution focuses on improving employee efficiency and eliminating assembly errors by providing instant, hands-free guidance during the trolley-packing process. It turns a reactive quality check into a proactive, real-time control system.

Key Deliverables: Lightweight AI Headsets: A proposal for inexpensive, single-user headsets that enable speech-to-text functionality to record packed items. Real-Time Feedback Logic (App Code): Logic that compares the employee's spoken input (e.g., "adding three beers, one wine") against the specific flight's digital specification. Error Detection and Guidance: The system uses the speaker/headset to immediately provide an audio alert or reminder if an item is missing or incorrect ("Alert: you are missing one white wine"). This prevents errors before the trolley is sealed and saves valuable rework time.

Challenges we ran into and Accomplishments that we're proud of

We ran into a couple of challenges while working with the speech-to-text model that goes into the headset for live recognition and feedback. The technologies and components we are using were giving us a hard time at the beginning of the set up but we managed to find a solution and we made it work. Overall, we're extremely proud of the final result of the project since it's exactly what we envisioned in the planning stage.