Check out our customer and restaurant-facing website here!

If you want to log in to a restaurant with populated data, you can use the following credentials.

Username: capstone1@gmail.com
Password: 1234

Concept

A smart table mat that will efficiently manage a restaurant.

Prompt Service

Waiter

Each table mat can detect when a guest has finished his/her meal and summon waiters. The mat uses light and weight sensors to detect this.

Waste Management

Waste Management

Each table mat detects and logs the amount of food left over on a plate. This data from table mats can be used to calculate the most/least popular dishes and food wastage

Crowd Control

Crowd

The number of active table mats can be used to determine the number of people at a restaurant. This information can ued to provide customers with real-time table-availability at a restaurant.

Motivation

Crowded Restaurant

Today, restaurants struggle to automate waiting of tables, have trouble knowing whether customers were satisfied with the dishes, and are unable to identify items from the menu that customers do not finish.

From the public perspective, avoiding waiting lines at restaurants still remains a challenge because we cannot find out in real-time how packed a restaurant is.

Therefore, a smart table mat will help restaurants automate prompt service, waste management and crowd control.

Competitive Analysis

Alternative methods to gather customer data/automate table-waiting


KFC

KFC

Baidu and KFC have partnered to use facial recognition to gather data about their customers such as age, gender and expression and recommend dishes. This does keep track of customers but only recommends based on past orders and does not detect accurately. It also doesn't work for one-time customers.

Airport

LGA

Many restaurants such as Eatsa in San Francisco and some restaurants in the New York airport let customers order food using tablets. They reduce the wait and ordering time but are extremely expensive and do not provide food-wastage analytics.

Requirements

Functional

  • Plate Detection
    • Detects the presence of a customer and hence the headcount in the restaurant
    • Calibrates the plate for accurate food wastage statistics (as plates differ from restaurant to restaurant)
  • Weight Sensing
    • Detect when customer is close to finishing a course and alert the waiters
    • Measure weight of food left over from each course
  • Processing
    • Poll every 20 seconds to detect if a plate has arrived. Until plate arrives, put itself (atmega) and the HM-10 to sleep to save power.
    • When a plate is detected:
      • Allows the plate to reach a steady-state
      • Poll ADC values of the FSRs every 5 seconds
      • The HM-10 will read these values
    • When the plate is removed go back to the first step
  • Low Power Communication
    • Transmit FSR values using BLE to an aggregator which keeps track of meal sessions
  • Aggregator
    • Keeps track of state and meal sessions
    • Chooses when to send information to the server (meal in progress) and when to wait for a session (customer not arrived)
    • Receives raw values over bluetooth and sends them over wifi to the server
  • Analytics for the Restaurant
    • Number of dishes ordered at different times of the day
    • Best and worst dishes for different meal times
    • Shortest and longest times for serving and consuming dishes
    • Food that is wasted the most and least
    • Menu items that bring in the most and least revenue
    • The most and least crowded times
  • Analytics for the Customer
    • Seat-availability in a restaurant based on table size
    • Relative popularity of dishes at a particular restaurant
    • Finding restaurants based on cuisine and price

Non Functional

  • Durable and Spill-Proof: Table-mat should not be repaired if customer spills food or beverages onto it
  • Non-obstructive: Nothing obstructing the area where the cutlery is kept
  • Real-time: customer-facing web application should update information by the minute
  • Reliable: Table-mat should not call waiters at the wrong time and not fail while customer is dining
  • Failure resiliency: In case of our components running out of power and not accepting connections our aggregator should take care of this and not crash.
  • Accurate: Server should maintain state from course to course in order to reflect real time analysis and report accurate weight readings
  • Easy Installation: Table-mats must be able to be installed and maintained without much effort (Easy power connections/long-lasting batteries to prevent changing frequently)

Use Cases

Restaurant



Customer


User Interaction

System Architecture

Components

Pika

Team

Hari

Hari

I think our project has potential to grow as it is easy to use and install and can provide a ton of beneficial analytical information. I love building things and designing PCBs!






hguduru@andrew.cmu.edu

Alina

Alina

Two fun facts about me:

Unofficial ambassador of Mountain Dew
I want to work with embedded systems to help humanity
Rule breaker


arath@andrew.cmu.edu

Mark

Mark

I once built a Daft Punk helmet. Currently interested in GPU computing and building embedded devices.









msfernan@andrew.cmu.edu

Varun

Varun

I like to eat good food, train for marathons and hike. I also love to code in C and java.




pvarun@andrew.cmu.edu