Robot Embodied Intelligence

Beijing University of Chemical Technology

Syllabus for Robot Embodied Intelligence  _ 

Ⅰ. General Information

Ⅱ.Curriculum Nature and Course Objectives

The course is mainly aimed at deepening and expanding the professional course learning of students majoring in robotics engineering. It is a technical, guiding, expandable, and practical course. Students need to understand and learn the concept of embodied intelligence, related technologies, and the application of embodied intelligence in the design of fully autonomous robots. Embodied intelligence refers to intelligent agents (such as robots, drones, smart cars, etc.) that interact with the environment in real time through physical entities, achieving integration of perception, cognition, decision-making, and action. It has been included in government work reports and as an important component of future industries, it has entered national strategic planning. Therefore, it is necessary to enable students to understand and master the latest cutting-edge concepts and technologies of robotics, in order to prepare for becoming a robotics engineering and technical talent. Through the study of this elective course, students are required to achieve the following goals:

G1: Understand the main research areas and development trends of embodied intelligence and fully autonomous robots, master the general basic knowledge of embodied intelligence and its development and related applications, cultivate learning interest, stimulate exploration motivation in the related fields of embodied intelligence robots, and provide reference for achieving future learning and development goals.

G2: Understand and master the basics of computation, perception, positioning, planning, and control of embodied intelligent robots; Master the application of large models in embodied intelligent robots; Understand the optimization methods for embodied intelligent robot systems, including algorithm safety and system reliability.

G3: Understand the application cases of embodied intelligent robots, comprehend and master the comprehensive design methods of embodied intelligent robots.

Ideological and political goals: To stimulate interest in learning and cultivate patriotism. Deepen the scientific spirit and professional ethics education of striving for excellence. Cultivate the spirit of innovation and craftsmanship, and strengthen professional education..

Ⅲ. The Corresponding Relationship between Course Objectives and Graduation Requirements

Table 1 Correspondence between course objectives and graduation requirements

Ⅳ. Teaching Contents and Requirements for the Lecturing Part

4.1 Background, history, and future of embodied intelligent robots (4 class hours)

4.1.1 Teaching Objectives (G1)

4.1.2 Teaching Content

(1) The impact and development of embodied intelligent robots

Understand the industry development overview and domestic and international development status of embodied intelligent robots.

(2) The problems and challenges of embodied intelligent robots

Understand the uncertainty of the application scenarios of embodied intelligent robots, including the cost of the embodied intelligent robot industry chain, the difficulty of system integration, data bottlenecks, and ethical standards.

(3) The history and future of embodied intelligent robots

Understand what embodied intelligence is, its development history, and traditional technological directions; The development relationship between embodied intelligence and artificial intelligence includes behavior based artificial intelligence, neurobiologically inspired artificial intelligence, cognitive developmental robotics, and physical embodiment and interaction.

(4) Tailored Intelligence Technology Based on Large Models

Understand the basic model classification of empowering embodied intelligent robots and the automation of embodied intelligent robot design.

4.1.3 Teaching requirements: Understand the history of embodied intelligent robots, comprehend the current development status of embodied intelligent robots, and master the existing problems and challenges of embodied intelligent robots.

Point of entry for ideological and political education: The origin and development of automatic machines. Introduce pictures, videos, principles, and stories of famous automatic machines in Chinese history and historical stories, such as wooden oxen and horses, seismometers, guide carts, and drum carts. Establish confidence, patriotism, and professionalism in Chinese history and Chinese manufacturing.

4.2 embodied intelligent robot computing, perception, and positioning system (8 class hours)

4.2.1 Teaching Objectives (G2)

4.2.2 Teaching Content

(1) Robot computing system

Understand and master the main content of autonomous robot computing systems, such as autonomous driving; Master the computing system of embodied intelligent robots.

(2) Perception System of Autonomous Robots

Understand and master the application of autonomous robot perception systems in object detection, semantic segmentation, stereo vision and optical flow, stereo vision and depth estimation; Understand the concept and main content of bird's-eye view perception, including LiDAR based BEV perception, camera based BEV perception, and fusion based BEV perception.

(3) Robot positioning system

Learn the positioning tasks, positioning principles, positioning algorithm principles, and positioning computing systems of autonomous robots; Understand multi-sensor data alignment and build a computing platform for autonomous robot positioning.

4.2.3 Teaching requirements: Understand the development trends of embodied intelligent robot computing, perception, and positioning systems, comprehend the main content of computing, perception, and positioning systems, and master the relevant technologies of computing, perception, and positioning systems.

Ideological and political entry point 2: Structural analysis of the extravehicular robotic arm of the Chinese space station. Introduce pictures and related principle introduction videos of the extravehicular robotic arm of China's space station. Establish confidence, patriotism, and professionalism in Chinese manufacturing.

4.3 Planning and Control System for Tailored Intelligent Robots (8 hours)

4.3.1 Teaching Objectives (G2)

4.3.2 Teaching Content

(1) Path planning and trajectory planning

Understand and learn path planning and trajectory planning methods, as well as variational methods, graph search methods, and incremental search strategies;

(2) Planning and Control Based on Reinforcement Learning

Learn the principles and procedures of reinforcement learning algorithms, and master the planning and control methods of embodied intelligent robots based on reinforcement learning.

4.3.3 Teaching requirements: Understand the concept of planning and control systems for embodied intelligent robots, and comprehend the planning and control methods for typical embodied intelligent robots; Master reinforcement learning algorithms and planning and control methods for embodied intelligent robots based on reinforcement learning.

4.4 Large scale model of embodied intelligent robot (8 class hours)

4.4.1 Teaching Objectives (G2)

4.4.2 Teaching Content

(1) Overview of ChatGPT for Robotics

Understand the background and work motivation of ChatGPT for Robotics, the outstanding ability of ChatGPT to solve robot control problems, the design principles and workflow of ChatGPT for Robotics, as well as its contributions and limitations.

(2) Application of Robotic Transformers Multimodal Large Model

(3) More modalities

4.4.3 Teaching requirements: Understand the concept of ChatGPT for Robotics, master the design principles and workflow of ChatGPT for Robotics.

Point 3 of ideological and political education: Application of brain computer interface technology in the field of embodied intelligent robots. Introduce pictures and instructional videos of the first clinical study case of implantable brain computer interface in China. Establish confidence in Chinese scientific research and Chinese manufacturing, understand the benefits of technology for humanity, and enhance students' professional learning motivation.

4.5 Building a Tailored Intelligent Basic Model (8 hours)

4.5.1 Teaching Objectives (G2)

4.5.2 Teaching Content

(1) Overview of Building a Tailored Intelligent Basic Model

Master background knowledge, meta learning, contextual learning, model pre training, and model fine-tuning.

(2) Key choices and trade-offs of pros and cons

(3) Overcoming computational and memory bottlenecks

4.5.3 Teaching requirements: Understand the construction principles of embodied intelligence basic models and master the construction methods of embodied intelligence basic models.

4.6 Accelerated Robot Calculation (2 class hours)

4.6.1 Teaching Objectives (G2)

4.6.2 Teaching Content

(1) Overview of Accelerated Robot Computing

(2) Robot positioning module acceleration

(3) Robot planning module acceleration

(4) Robot control module acceleration

(5) Factor diagram: a universal template for robot accelerators

4.6.3 Teaching requirements: Understand the calculation principles of accelerating robots, master the acceleration methods of robot positioning modules, planning modules, control modules, and the general template for robot accelerators.

4.7 Algorithm Security (2 class hours)

4.7.1 Teaching Objectives (G2)

4.7.2 Teaching Content

(1) Overview of Algorithm Security

(2) Artificial intelligence security: a stumbling block between algorithms and applications

(3) Attack and Defense of Deep Neural Networks

Understand various attacks, including escape attacks, poisoning attacks, exploration attacks, and defense methods.

(4) Security Issues in Large Models

(5) Big model security risks vs. embodied intelligent robot security

4.7 System Reliability and Data Challenges (2 credit hours)

4.7.1 Teaching Objectives (G2)

4.7.2 Teaching Content

(1) Reliability vulnerabilities in robot systems

Understand the reliability vulnerabilities of the robot body, the reliability vulnerabilities of the robot computing system, and common methods to improve system robustness, such as adaptive redundancy methods, to enhance robustness while reducing system burden.

(2) The data challenge of embodied intelligence

Understand the data value and data bottlenecks of embodied intelligence

4.7.3 Teaching requirements: Understand the reliability and data system concepts of embodied intelligent robot systems, and master common methods to improve system robustness.

4.8 Application cases of embodied intelligent robots (6 class hours)

4.8.1 Teaching Objectives (G3)

4.8.2 Teaching Content

(1) Application cases of embodied intelligent robots

4.8.3 Teaching requirements: Understand various application cases of embodied intelligent robots and master the design methods of embodied intelligent robots.

Ⅴ.Teaching Contents and Requirements for the Practical Part

No

Ⅵ. Evaluation Standards

The assessment methods of this course include process assessment and final assessment. The process assessment includes class performance (including attendance, questioning, answering questions, group discussion) and class test.

The final assessment of the course is a comprehensive assessment of the process assessment and the final assessment, and based on this, the achievement of the course objectives is evaluated.

Ⅶ.Textbooks and Recommended References

7.1 Teaching materials

 [1] Gan Yiming et al., Embodied Intelligent Robot System, Electronic Industry Press, First Edition, 2024

7.2 Reference books

[1] Liu Ruochen, Introduction to Artificial Intelligence, Tsinghua University Press, 2021

 [2] Zhang Weinan, Shen Jian, Yu Yong et al. Hands-on Learning and Reinforcement Learning, Posts and Telecommunications Press, 2021

Written by Guohua Chen