Flight Testing and Fundamentals of Model Predictive Control (MPC) for Quadrupedal Locomotion Summer School
1. Project Title & Introduction
· Project Title: Fundamentals of Model Predictive Control (MPC) for Quadrupedal Locomotion（四足机器人模型预测控制基础）
· Project Introduction: This course provides a specialized deep dive into MPC, the cornerstone of modern dynamic locomotion in quadrupedal robotics. The program covers the entire pipeline of MPC-based control, from simplifying multi-body dynamics into Single Rigid Body Models (SRBM) to formulating and solving constrained Quadratic Programming (QP) problems in real-time. Students will explore how MPC predicts future states to plan optimal ground reaction forces, enabling robots to maintain balance and achieve robust gait patterns under various disturbances.
- Course Structure
· Module I: Foundations of Predictive Control
o Discrete-time state-space representations.
o The concept of the receding horizon and cost function design.
· Module II: Dynamic Modeling for MPC
o Simplification techniques: From full dynamics to the Single Rigid Body Model (SRBM).
o Linearization and discretization of robot dynamics for real-time optimization.
· Module III: Optimization & Constraints
o Formulating the Locomotion QP problem.
o Incorporating physical constraints: Friction cones, motor torque limits, and unilateral contact forces.
· Module IV: Implementation & Simulation
o Case studies: Implementing Trot gaits in physics engines.
-Instructor
· Dongyi Li
-Student Qualifications
· Academic Level: Open to all academic levels (Undergraduate, Master’s, and Ph.D. students).
· Academic Requirements: Applicants should have a foundation in Control Theory (e.g., PID, State-space representation) and a basic understanding of Robotics Kinematics. Familiarity with programming (C++ or Python) and simulation environments (such as Gazebo or Mujoco) is highly recommended to fully benefit from the technical sessions.
2. Project Title & Introduction
· Project Title: Trustworthy LLMs: A Road to Reliable AI Technology (可信大语言模型：通往可靠人工智能之路）
· Project Introduction: This talk provides a comprehensive introduction to the foundational principles and covers the critical dimensions of Trustworthy Large Language Models (LLMs), including fairness, robustness, safety, reliability and explainability. Participants will explore how we move beyond “black-box” predictions toward creating AI systems that are not only powerful but also transparent, predictable, and ethically grounded. This session aims to provide a clear roadmap of the current trustworthy AI landscape.
-Course Structure
· Module I: Introduction of Large Language Models
o An evolution process of language models.
o Key pillars of the trustworthy AI framework.
· Module II: Fairness and Robustness
o Promoting fairness: Identifying and mitigating social biases and disparate performance across different groups.
o Enhancing model robustness: Ensuring stability against prompt attacks, poisoning, and unexpected distribution shifts.
· Module III: Safety and Reliability
o Upholding safety standards: Preventing the generation of harmful content, privacy violations, and unlawful conduct.
o Ensuring system reliability: Reducing hallucinations and miscalibration to provide consistent and factually accurate outputs.
· Module IV: Implementation and Deployment
o Case studies: Successful examples of reliable AI enhancing human decision-making.
o Real-world deployment: Integrating trustworthy LLMs into critical domains.
-Instructor
· Yingjie Li
-Student Qualifications
· Academic Level: Open to all academic levels (Undergraduate, Master’s, and Ph.D. students).
3. Project Title & Introduction
· Project Title: Basics of Metamaterials（超材料基础）
· Project Introduction: This course mainly teaches the basic concepts, classifications, different functions, and design methods of metamaterials, enabling students to establish a preliminary and comprehensive understanding of metamaterials, master the basic methods of metamaterial design, preparation, and testing, and at the same time understand the latest development trends and academic frontier achievements of metamaterials in different fields such as statics, waves, and heat transfer.
-Course Structure
· Module I: Introduction
o Understand the concept, classification, characteristics, application scenarios, development history, etc. of metamaterials.
· Module II: Mechanical Metamaterials
o Understand mechanical concepts such as negative Poisson's ratio, energy absorption efficiency, and nonlinearity.
o Master different types and characteristics of mechanical metamaterials such as bending-dominated and compression-dominated types.
o Master the basic principles of unit cells, periodic boundaries, and finite element simulation.
o Master the main design methods of metamaterials such as negative Poisson's ratio, high stiffness/strength, high energy absorption, and isotropy.
· Module III: Chiral Metamaterials
o Understand the concept, characteristics, judgment methods and common chiral structures of chirality.
o Master the design method of pressure-torsion effect based on chiral metamaterials.
o Understand other unconventional functions realized by chiral metamaterials.
· Module IV: Zero-Mode Metamaterials
o Master the general elastic relationships, modulus boundaries, and characteristics of metamaterials.
o Understand the concept of pentamode, the properties and application prospects of pentamode metamaterials.
o Master the similarities and differences between pentamode materials and liquids through comparison.
o Understand common configurations of pentamode metamaterials.
· Module V: Acoustic Metamaterials
o Master basic concepts such as band gap, phonon, and vibration, as well as acoustic wave control equations, etc.
o Understand the general spring-mass equivalent model for acoustic wave transmission and the main control methods of acoustic waves.
o Understand the typical applications and main functions of acoustic metamaterials.·
· Module VI: Nonlocal Metamaterials
o Master basic concepts such as non-local effects, Roton dispersion relations, and nearest-neighbor and next-nearest-neighbor interactions.
o Understand the unconventional functions such as multiple scattering caused by non-local effects and their application prospects.
o Learn about non-local interaction models in the fields of electromagnetics, acoustics, and mechanics, and grasp their similarities and unique features through comparison.
-Instructor
· Muamer Kadic, Xiaojun Tan
Student Qualifications
· Academic Level: Open to all academic levels (Undergraduate, Master’s, and Ph.D. students).
Contact Us
Professor Xiaojun TAN, xiaojun_tan1@163.com
