Real-Time Task Planning Algorithm for Home Care Robots Based on Transformer Architecture

Global population aging strains social security systems. Seniors prefer aging in place, but face safety risks. Smart eldercare ecosystems leverage AI and IoT for health monitoring. This study proposes a real-time task planning algorithm for home care robots using a dynamic Hamiltonian graph model. Time-sensitive priority functions ensure high-risk response. An enhanced Transformer architecture fuses spatial-priority features in encoders and generates dynamic sequences via spatiotemporal attention in decoders. An asynchronous dual-network reinforcement learning framework optimizes path costs. Experiments demonstrate effective prioritization of urgent tasks and rapid adaptation to dynamic perturbations while maintaining execution efficiency.