This assistive feeding robot integrates speech, vision, and depth sensing to help elderly or mobility-impaired users eat safely with minimal caregiver support. It features spillage risk prediction and real-time motion adjustment. Users interact via voice or touchscreen. The modular, scalable system suits home care, long-term care, and smart healthcare environments.

This technique combines a guided diffusion policy with a food spillage predictor, overcoming the limitation of conventional methods, which lacks real-time adaptability. It introduces a novel approach by embedding a spillage-aware objective into the denoising process and training the predictor using simulated data. This enhances the controllability and safety of trajectory generation, with its generalization ability validated in real-world food manipulation tasks.

With the rapid aging of society and growing shortage of caregiving personnel, assistive feeding robots hold significant industrial application potential. This system enhances the feeding autonomy of elderly and disabled individuals, reduces caregiver workload, and can be deployed in long-term care facilities, home care, and smart healthcare markets. Its modular architecture facilitates integration with other care technologies, supporting commercialization and scalable production. It aligns with Taiwan’s trends in smart healthcare and aging technology innovation.