Mobile Kiosk Task Orchestration for Offline Gaming Venues

The patent implements a universal mobile kiosk platform that can dynamically load and execute multiple different applications from cached datasets. Instead of requiring separate dedicated hardware for each task, a single kiosk unit can be configured to perform various functions (customer service, gaming, information display, etc.) by loading appropriate application code and data from its local storage. This multi-functionality approach allows one device to replace multiple specialized devices, reducing overall system complexity while maintaining high versatility.

The system employs dynamic application loading and execution capabilities, where the mobile kiosk can switch between different applications based on task requirements. The kiosk controller dynamically loads application code and associated datasets from cached storage into memory, executes the required application, and switches to other applications as needed. This dynamic behavior allows the system to adapt its functionality in real-time without requiring all applications to be simultaneously active in memory, thereby managing device complexity effectively.

The system implements periodic task scheduling and execution strategies where computationally intensive machine learning tasks are performed in periodic batches rather than continuously. The kiosk controller monitors energy levels and schedules heavy computational tasks during periods when energy availability permits, interspersing them with lower-power operations. This periodic execution pattern allows the system to accumulate and expend energy in manageable cycles, extending overall operational duration while still delivering required computational power when needed.

The patent employs adaptive parameter adjustment for machine learning model execution, where the system dynamically selects and executes models with computational complexity matched to available energy resources. When energy levels are high, the system can execute more computationally intensive models with higher accuracy. When energy levels drop, the system automatically switches to lighter-weight models or reduces model execution frequency. This parameter adaptation allows the kiosk to maintain useful machine learning functionality across the full range of battery charge states, effectively extending operational duration without completely sacrificing computational capability.