In 2023 (the 16th) China College Students Computer Design Competition, the Intelligent Service Robot Practice Competition was officially launched. Dataa Robotics, the Organizing Committee of the China College Students Computer Design Competition and Peking University once again jointly organized this competition.The competition is co-sponsored by the Ministry of Education's computer majors, software engineering major teaching steering committee and other units. As a leading unicorn enterprise in the field of smart robots, Dataa Robotics is the world's leading creator, manufacturer and operator of cloud robots, together with Peking University, Dataa Robotics initiated and hosted the service robot practice competitions in 2021 and 2022.
The competition has four major highlights: robot metaverse, HARIX open ecology, multi-modal embodied AI, and humanoid robots. Based on the HARIX OS and HARIX RDK robot development kits, the contestants realized various algorithm training and program verification of the humanoid robot Cloud Ginger XR-1 in simulated real scenes, and additionally achieved the transformation from virtual to real and the combination between the two. The intelligent migration enables the physical robot to complete the work in the real scene.
1.Robot Metaverse
"Robot Metaverse" is an extremely realistic simulation environment that digitizes the physical world in real time. It is also a working and learning environment for robots to understand the physical world, intelligent training, and virtual-real interaction. This enables robots to synchronize with the work of the physical world through digital twins, forming a robot metaverse platform that can carry "human intelligence + artificial intelligence".
2.HARIX Open Ecology
HARIX RDK supports the import of third-party algorithms, covering various algorithms such as voice, vision, navigation, and motion control. Users only need to use the open standard interface of HARIX OS to realize dynamic registration of algorithm skills, acquisition of robot sensor data and status, and algorithm result feedback, etc., also achieve resource co-construction and sharing, including data models of physical world scene, robot digital twin model, artificial intelligence algorithm, industry application program, etc. In this way HARIX forms a global and active UGC community ecology.
3.Embodied Intelligence of Multimodal AI and Migration of Sim2Real
The technological development of artificial intelligence is changing from "Internet AI" based on internet static data sets (such as ImageNet, COCO, VQA) to "Embodied AI", which means intelligent Robot can do multi-modal interactive action in real environment. The embodied AI represented by intelligent robots will gather multi-modal AI capabilities to promote the paradigm shift in the development of artificial intelligence technology and new breakthroughs in the application field. After the robot skills are trained in the simulation environment, the HARIX cloud brain completes the seamless intelligent migration from the virtual environment to the physical scene, empowering the physical robot to complete the work in the real world.
4.Humanoid robot
Cloud Ginger XR-1, a cloud-based humanoid intelligent service robot with 34 compliant joint actuators, can flexibly control the whole body with precise force feedback, and the dexterous hand can perform flexible grasping, combined with various visual, auditory, tactile and other sensors (2D camera, 3D camera, Ultrasonic radar, inertial navigation, lidar, microphone array, etc.). Connected to the cloud brain through 4G/5G modules, Ginger can complete various complex scene services under the processing of the HARIX OS distributed operating system.
This year's competition set up two tracks: robot cafe waiter and robot playing billiards, to examine the comprehensive ability of contestants in robot natural speech understanding, visual recognition, navigation, walking, and item interaction operation. Through multi-modal algorithm training and application development, the Ginger XR-1 robot can be trained to be a qualified waiter or pool player.
Track 1: Robot Cafe Waiter
Based on the interactive operation tasks of the intelligent service robot in the cafe, the contestants trained the Ginger XR-1 robot to be a qualified cafe waiter through multi-modal algorithm training and application development.
Challenge task A: Use the coffee machine to make coffee and send it to the customer who ordered
Challenge task B: Ginger robot places chairs
Challenge task C: Ginger operates smart devices
Open task: Open task based on the cafe scene: In this scene, the open task design is carried out, and the task requires interactive operation of items in the simulated cafe scene. For visual recognition, please consider the accuracy and optimization of the algorithm model. For the interactive operation between robot and item, please refer to the implementation method of the interactive operation of the existing model.
Track 2: Robot playing billiards
The interactive operation task is the intelligent service robot playing billiards in the billiards arena, in which players complete the robot's fixed-point hitting task based on the HARIX OS operating system and the digital twin simulation environment constructed in accordance with the standards of the billiards arena.
Challenge task: Fixed-point hitting
Place the cue ball at a fixed position on the table, and place the target ball at the other fixed position. Every time a hit is completed, the cue ball and the target ball are reset, and the task of hitting the ball is completed within the specified time. The fixed-point hitting task is dismantled into the following three subtasks:
Subtask 1: The robot can hit the cue ball accurately, and the cue ball can hit the edge of the table
Subtask 2: The robot can hit the target ball by hitting the cue ball
Subtask 3: The robot can hit the target ball and enter the hole by hitting the cue ball
Open task: an open task based on the scene of playing billiards
According to the provided API list of blueprint skill sets and the open HARIX algorithm API interface, participating teachers and students performed task design and algorithm implementation based on the scene of the robot playing billiards.Game scheduleAfter the participating teams pass the school selection, they will participate in the preliminaries of the "Intelligent Service Robot Practice Competition" in the direct report competition area, and then be promoted to the national competition.
Dataa Robotics' WeChat Account
