Walkie 2

Robot

GitHub

An intelligent autonomous service robot, our EIC's friend, is built to participate in Robocup@home 2022 competition. We initiated Walkie in spring 2021 with a strong belief that Robots are good human companions. Walkie can do various household tasks and be a party host. Things he can do are picking and placing the object on the shelf, tidying up the garbage on the floor, serving the meal up to the customer's order, helping hold a bag, and following a user.

Walkie

Walkie is a domestic service robot that helps you to do house chores and be your great companion.

Walkie

We might think that our daily routine is boring and simple. But, for a robot like Walkie, a simple task may require much complex planning and motion.

Walkie

Walkie can arrange groceries in the cabinet, take out the trash, carry your bag, serve the cool drink from the refrigerator, and communicate with you in English.

Walkie

Artificial Intelligence plays a significant role in making Walkie reliable and making intelligent decisions in doing boring human routines.

Our team, EIC from Chulalongkorn University, is developing Walkie's body and brain to make our robot not just a machine. He is our friend.

Team Description Paper for Robocup@Home 2023 OPL

Tasks

Restaurant

One of the most recurring tasks in Robocup@Home. Appearing during the semi-final.

This task tests the robot's interactivity with the synergy of navigation, manipulation, NLP, and Computer Vision.

Simulating a restaurant, the robot is placed in the dining area looking for a customer to usher. When the customer raised their hand, the robot walks over and takes requests from the customer whether it is taking an order, getting the bill, etc. The robot then serves the customer their food and begins the process all over again.

Receptionist

Walkie successfully introduces a newly come guest to the host (person sitting in the room). Using speech recognition and face recognition, Walkie is able to tell the guest`'s name and favorite drink.

Hand me that

Another second round of tasks in RoboCup@Home. This task focuses on Object detection and Human Robot Interaction. The robot must identify the object that the human requires. Objects are randomly group together between 3 - 5 items. The robot must identify the operator(human) and wait for their instruction.

Features

Instance Segmentation

Object detection and instance segmentation using YOLOv8 trained on a custom dataset. This video depicts a robot wandering around an office with random objects and furniture. This computer vision model can detect those objects and return their segmentation mask.

3D Self-Mapping

Walkie2 supports navigating in an unknown area, the grid map is being created in real-time while the robot is operating. Moreover, the data from the depth camera is used to create a voxel layer for 3D obstacle avoidance. In this video, the robot follows a target while trying to avoid any obstacles in its path.

Human Posture Detection

This video demonstrates the hand raising detection this feature is used in tasks for Robocup@Home. The Restaurant(hyperlink) task is one example, hand raising detection is used to usher the robot to the customer.

The hand raising detection is a subset of Posture detection, the one used in the Restaurant task is the older version. The new detection, it's able to detect joints faster, and more accurately, and able to differentiate people in crowded areas better