You are on page 1of 13

Team GMT3

AA274 Final Project Presentation


Autumn 2020
Lydia Chan, Clara Keng, Natasha Ong, David Yue
2020 Project Challenge:

➢ Autonomous food
delivery
➢ Navigating city without
running into obstacles
➢ Receive and process
requests
➢ Collect food from
vendors and make
deliveries
Meet Nevarizona
though slightly slow-moving at times,

Packed with powerful features:


➢ Advanced AI image
recognition
➢ Autonomous navigation
➢ And much much MORE!

3
High-Level Controls
Our robot tackles the
biphasic task at hand with
different acutely-tailored
logic for each process.

Our goal is to bring accurate


and efficient results in both
the exploration and the
delivery phases.
Software Stack

● detector_mobilenet
○ detect the objects from camera
● navigator
○ encapsulate the FSM logic
● goal_commander
○ publish cmd_nav
● landmark_map
○ track and log the objects in the map
○ publishes goals for delivery phase
● marker_pub
○ publish the markers into RViz
● ...and other supporting infrastructure
Extension 1: Stop Signs

Following traffic regulations is crucial


in real-world scenarios, especially food
delivery.

Our convolutional neural network


(CNN) detects stop signs, and our robot
makes the necessary stop. Only when it
is safe to do so does it cross the road.

Safety comes first!


Extension 2: Shortest Path - Genetic Algorithm Heuristic
for Solving Traveling Salesman Problem
Finding the shortest path between a
varying number of locations is not an
easy problem. In fact, it is an non-
deterministic polynomial time (NP-hard)
problem.

We utilize a genetic algorithm to


heuristically optimize local minima to
find a satisfactory short path that
traverses the necessary locations.
Extension 3: Zoological Salutations
With the friendliness of a local park
ranger, our robot visits animals and
makes the correct greetings!

A “meow” for a cat, a “woof” for a dog,


a “hello” for a person. We know that all
of our greetings will be understood by
the cute animalia in their native
languages.
Extension 4: Finding TA’s Houses: Multi-Channel
Surjective Food Delivery Process
In a time of COVID, we can’t be
walking around to visit friends
ourselves! Nevarizona notes the
locations of our friends (our TAs) as it
explores the world, and offers us the
option to have it visit them on our
behalf!

We have made the many-to-one problem


into a many-to-many situation. This
way, we can also send food to our
friends. You know it’s dinner time!!
Extension 5: Mark and label objects on map!
We prize organized bookkeeping, so we
publish the corresponding markers for
all our vendors and friends!

A sparkly pink denotes the location of


each of our vendors, our TA friends, and
even the neighborhood stop sign! We
believe that this information (within
confidentiality bounds, of course!) will
be shared to advance our delivery
mission!
Extension 6: Enthusiasm for the job!
We celebrate each pick-up, since we are
super happy for our very wonderful
vendors!

Instead plain ol’ idling at the curbside,


we do a nice lil spin as we wait for the
food! We’re just so EXCITED to pick
up the food and deliver it to our friends!

Let’s do a little happy dance!


Extension 7: 24/7 Delivery + Tech Support (jk)
4 AM and need to satisfy your
munchies? We’re here for you!

Our team spans 4 time zones and 2


continents! With full coverage, we will
always be there to deliver your food or
offer tech support.

The sun does not set on Genbu!


Time to meet
Nevarizona!

13

You might also like