Robotic meal-bowl assembly and fresh-ingredient portioning
This is the core contact-rich task inside Aramark's deployed RoboEatz ARK system: a robotic arm reaches into storage bins to retrieve and portion fresh ingredients, then assembles them into customized bowl meals. The system draws from up to 80 fresh ingredients and prepares up to four personalized bowls at a time within a 400-sq-ft footprint, managing food storage, meal preparation, plating and cleaning end-to-end. The objects are highly heterogeneous and difficult: delicate leafy greens that tear and wilt, soft proteins, grains and starches that clump, slippery sauces, and produce of variable size and shape. The task sits at the point of service in 24/7 healthcare dining, where it must hit nutritional/portion compliance for specific dietary needs, so a mishandled or wrong-weight scoop is a quality failure, not just a re-grip. What makes it hard for a robot is exactly what AGD targets: grasping/scooping items of unpredictable geometry and compliance without crushing, smearing, or dropping them, and confirming a successful, correctly-portioned pickup that vision alone struggles to verify. We identified this through our own research; we have not confirmed the specifics with the customer directly. This page is our researched read — a starting point for that conversation.
What the task is
RESEARCHED · our reconstructionThis is the core contact-rich task inside Aramark's deployed RoboEatz ARK system: a robotic arm reaches into storage bins to retrieve and portion fresh ingredients, then assembles them into customized bowl meals. The system draws from up to 80 fresh ingredients and prepares up to four personalized bowls at a time within a 400-sq-ft footprint, managing food storage, meal preparation, plating and cleaning end-to-end. The objects are highly heterogeneous and difficult: delicate leafy greens that tear and wilt, soft proteins, grains and starches that clump, slippery sauces, and produce of variable size and shape. The task sits at the point of service in 24/7 healthcare dining, where it must hit nutritional/portion compliance for specific dietary needs, so a mishandled or wrong-weight scoop is a quality failure, not just a re-grip. What makes it hard for a robot is exactly what AGD targets: grasping/scooping items of unpredictable geometry and compliance without crushing, smearing, or dropping them, and confirming a successful, correctly-portioned pickup that vision alone struggles to verify.
Is this the actual task and sequence? What are the real tolerances, cycle rate, and reject criteria, and which steps are today's manual bottleneck? Answering these is what turns this from a researched signal into a validated use case.