Nutrition assistant robot uses graphs and language models for smarter advice
Nutri-ATLAS: Embodied Agent for Tabulated Lookup and Assistance for Smarter nutrition
Artificial Intelligence
Summary
Nutrition advice is tricky when food descriptions are unclear or nutrition data is missing. The authors created Nutri-ATLAS, a robot agent that understands food nutrition by linking information in tables (food and nutrient data) with language models and physical sensing. It can suggest ingredient substitutions, fill in missing nutrition details, and recommend meals based on what food is actually available. The system was tested on several tasks and datasets and showed it can follow dietary restrictions like allergies and health targets in the real world.
What this means in practice
- •For dietary assistance developers: Build robots that provide personalized nutrition advice adapting to real food environments and dietary needs.$Commercial implications: Enables commercial development of digital nutrition assistants that combine sensing and reasoning for customized meal planning and guidance.
- •For robotics platform engineers: Integrate graph-based nutrition reasoning with language models for robots that autonomously navigate kitchens and recommend meals.
Authors
Uttej Kallakuri, Boxun Hu, Ankur A. Butala, Najim Dehak, Tinoosh Mohsenin
Abstract
Generative and Agentic IoT systems offer a promising foundation for digital healthcare applications that combine sensing, personalized reasoning, and autonomous interaction in real-world environments. Nutrition assistance is a natural use case, but existing Large Language Model (LLM)-based systems are often limited to passive text interaction and static context, making them unreliable when food descriptions are ambiguous or nutritional evidence is missing. We propose Nutri-ATLAS, an Embodied Agent for Tabulated Lookup and Assistance for smarter nutrition in the real world. It integrates graph-grounded nutrition reasoning, hardware-aware LLM selection, and robot-based evidence acquisition. Nutri-ATLAS builds a unified Food-Nutrient knowledge graph from USDA FoodData Central and FoodKG and learns 64-dimensional GATv2 food and recipe embeddings. A shared hybrid graph-text scoring mechanism supports food nutrition extraction, nutritional gap filling, substitute retrieval, and recipe-level meal composition, while an LLM-guided skill interface navigates landmarks, updates dietary-context and food-accessibility memory, and grounds recommendations in observed food availability. We evaluate Nutri-ATLAS across nutrient estimation, substitution retrieval, recipe recommendation, patient-profile adherence, edge deployment, and real-world embodied execution. On HealthyFoodSubs, the hybrid retriever achieves 37.9% MAP, 80.7% RR@5, and 90.1% RR@10. On NutriBench v2, Dense+GAT retrieval grounds nutrient estimation across nine quantized Qwen3.5-9B configurations. On PFoodReQ, Nutri-ATLAS reaches 78.8% MAP, 83.0% MAR, and 77.5% F1. A patient-profile study shows adherence to allergy and healthy-target constraints for all selected cases.