Pill Guard

World Model for training.

Inspiration

What it does

How we built it

Challenges we ran into## Inspiration

My grandparents in China are dealing with this problem right now. Family members often have to remind them to take their medicine and physically get the pills out for them, because even with a Monday/Tuesday/Wednesday pill box, doses still get forgotten. The problem gets worse when family members are busy or unavailable. A home monitoring camera also feels intrusive and uncomfortable for them, which made us think about a more targeted and respectful solution.

What it does

Pill Guard is a robotic medication adherence assistant for older adults. It reads a medication schedule, uses a robot arm to press the correct dispenser control, and checks whether the pills were removed from the tray. If the medication is still there after 20 minutes, it sends an alert to a child or caregiver. The goal is to go beyond reminders and add verification plus escalation.

How we built it

We built Pill Guard as a pipeline with four parts: prescription understanding, scheduling, robotic actuation, and adherence monitoring. The prescription component extracts medication instructions and converts them into a structured dose event. The robot arm uses an onboard camera to localize the dispenser target and perform a fixed pushing motion. After dispensing, the same camera checks whether the pills remain in the tray and triggers an alert if the dose appears to be missed.

Challenges we ran into

The biggest challenge was scope control. A full pill-handling robot would require reliable manipulation of small, slippery pills, which was too risky for a hackathon. We narrowed the problem to robotic interaction with an existing dispenser and visual monitoring of the output. Another challenge was making accurate claims, since detecting that pills were removed is not the same as verifying they were swallowed.

Accomplishments that we're proud of

We are proud that Pill Guard demonstrates a complete closed-loop workflow instead of just a reminder or a single robotic motion. The project connects prescription understanding, robotic actuation, visual checking, and caregiver alerting into one system. We are also proud that we chose a realistic design that works with existing dispensers rather than requiring custom pharmacy hardware. That made the concept more practical and easier to demonstrate clearly.

What we learned

We learned that assistive robotics works best when the task is simplified and the workflow is tightly structured. It is more effective to make the robot perform a small number of reliable actions than to force it into a complex manipulation problem. We also learned that privacy matters: families may want help with adherence, but they do not necessarily want constant home surveillance. Finally, we learned to be precise about what our system can verify and what it cannot.

What's next for Pill Guard

The next step is to improve the robustness of dispenser-target recognition and tray-state detection in less controlled environments. We would also want to support more medication workflows and make caregiver notifications more useful and configurable. Beyond the hackathon, we would explore a less invasive home deployment model that preserves privacy while still enabling adherence monitoring. Long term, Pill Guard could become a practical assistive layer for families supporting older relatives remotely.