| NCHU Course Outline |
| Course Name | (中) 實體人工智慧(7851) | ||||
| (Eng.) | |||||
| Offering Dept | General | ||||
| Course Type | Elective | Credits | 3 | Teacher | Kuan-Chu Peng |
| Department | General | Language | English | Semester | 2026-FALL |
| Course Description | Recent years have witnessed remarkable breakthroughs in artificial intelligence—from large language models transforming how humans interact with information to generative models capable of creating increasingly realistic images, videos, and virtual worlds. As AI continues to advance, the next major frontier extends beyond the digital world: Physical AI, where intelligent systems perceive, reason, make decisions, and act in the physical world. Unlike AI systems operating primarily in digital spaces, Physical AI must confront the complexity and uncertainty of the physical world: incomplete observations, changing environments, embodiment constraints, sensor and actuator failures, interactions with people, and consequences associated with physical actions. This course introduces the foundations and recent advances of Physical AI through the lens of real-world development, evaluation, and deployment. Specifically, as AI moves into the physical world, we must continuously ask: What challenges emerge when the system leaves the benchmark and enters the real world? How should we evaluate a Physical AI system beyond predefined metrics and state-of-the-art performance? How can evaluation reveal system limitations and, more importantly, guide inference, system improvement, and future development? |
||||
| Prerequisites | self-directed learning in the course | Y | |||
| Relevance of Course Objectives and Core Learning Outcomes(%) | Teaching and Assessment Methods for Course Objectives | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Course Objectives | Competency Indicators | Ratio(%) | Teaching Methods | Assessment Methods | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 1. Explain the foundations and recent advances of Physical AI. 2. Describe and apply the Sense–Plan–Act paradigm to robotic navigation and manipulation, connecting sensing and perception with planning, decision-making, control, and action. 3. Identify real-world challenges in the development and deployment of Physical AI systems beyond predefined benchmarks, including uncertainty, environmental dynamics, embodiment constraints, distribution shifts, human interaction, safety, and reliability. 4. Recognize the critical role of evaluation beyond predefined metrics and state-of-the-art performance, and analyze system capabilities, limitations, failure modes, robustness, generalization, and safety under realistic conditions. 5. Design evaluation strategies for Physical AI systems that reflect real-world requirements, scenarios, and system behaviors rather than relying solely on aggregate benchmark metrics. 6. Apply evaluation insights to guide inference and system improvement, using observed failures, limitations, and operating conditions to improve Physical AI system design and behavior. |
|
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Course Content and Homework/Schedule/Tests Schedule | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Evaluation | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ● Computer Assignment: 50% - HW1: 3D Scene Reconstruction and Mapping - HW2: Robot Navigation - HW3: Robotics Manipulation ● Course Project: 50% The Course Project is designed to provide students with hands-on experience across the full lifecycle of Physical AI, from real-world data collection, to policy learning in simulation, and ultimately to evaluation. More details will be shared in the kick-off meeting. |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Textbook & other References | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Teaching Aids & Teacher's Website | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 本次作業(Computer Assignment)將使用 Habitat 虛擬器。作業內容主要以基本操作為主, 並不涉及高負載的模擬或模型訓練,因此整體硬體需求不高。 確認電腦是否符合最低需求時,可以優先檢查以下兩點: ● 電腦是否具備可用的顯示晶片(GPU) ● 記憶體(RAM)是否至少為 8 GB 原則上即可執行本次作業。 詳細硬體需求如下: GPU ● 最低:支援 OpenGL 3.3 / EGL,至少 2 GB VRAM ● 建議:4 GB VRAM 以上 CPU ● 最低:x86-64,4 個實體核心 ● 建議:8 核心以上,Boost 時脈 ≥ 3.5 GHz RAM ● 最低:8 GB ● 建議:16 GB 以上 儲存空間 ● 最低:20 GB 可用空間 ● 建議:30 GB 以上,並使用 NVMe / SATA SSD 作業系統 ● MacOS ● Linux ● Windows (必須安裝 WSL ) |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Office Hours | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Sustainable Development Goals, SDGs(Link URL) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Please respect the intellectual property rights and use the materials legally.Please respect gender equality. | |
| Update Date, year/month/day：2026/09/01 18:02:29 | Printed Date, year/month/day：2026 / 10 / 17 |
| The second-hand book website：http://www.myub.com.tw/ | |