Assessment Package: Introduction to Physical AI

Overview

Component	Weight	Format
Theory Quiz	15%	Multiple choice + short answer
Lab Assessment	35%	Jupyter notebooks with auto-grading
Simulation Project	35%	Code submission with test harness
Ethics Component	15%	Written reflection + peer review

Theory Quiz (15%)

Time Limit: 20 minutes Passing Score: 70% Attempts Allowed: 2

Questions

Q1 (Multiple Choice, 2 points)

Question: What distinguishes Physical AI from traditional software AI?

Options: a) Physical AI uses more computing power b) Physical AI must interact with the real world through sensors and actuators c) Physical AI is always humanoid in form d) Physical AI does not use machine learning

Correct Answer: b Explanation: Physical AI systems are defined by their need to perceive and act in the physical world, dealing with uncertainty, real-time constraints, and physical consequences that pure software AI does not face. Learning Objective: Define Physical AI and distinguish it from traditional robotics and software AI

---

Q2 (Multiple Choice, 2 points)

Question: The Embodiment Hypothesis suggests that:

Options: a) All intelligent robots must have human form b) Intelligence emerges from the interaction between brain, body, and environment c) Robots cannot be truly intelligent d) Physical form has no effect on cognitive capabilities

Correct Answer: b Explanation: The Embodiment Hypothesis is the principle that intelligent behavior cannot be fully separated from physical form—the body shapes cognition. Learning Objective: Explain the role of embodiment in artificial intelligence

---

Q3 (Multiple Choice, 2 points)

Question: Which generation of robots introduced programmable automation?

Options: a) First generation (1950s-1960s) b) Second generation (1970s-1980s) c) Third generation (1990s-2010s) d) Fourth generation (2020s+)

Correct Answer: a Explanation: First-generation industrial robots like Unimate introduced programmable automation to manufacturing, though they lacked sensors and operated in structured environments. Learning Objective: Describe the historical evolution from industrial robots to intelligent humanoids

---

Q4 (Short Answer, 4 points)

Question: List three key advantages of the humanoid form factor for robots intended to operate in human environments. Explain one in detail.

Rubric:

• 1 point for each valid advantage listed (max 3)
• 1 point for detailed explanation

Sample Answer:

1. Compatible with human-designed environments (doors, stairs, tools)
2. Intuitive human-robot interaction through familiar body language
3. Versatile manipulation with human-like hands Detailed: Human environments are designed for human bodies—standard door widths, stair heights, counter heights, and tool handles. A humanoid robot can navigate these spaces without environmental modification. Learning Objective: Identify key components and capabilities of modern humanoid robot systems

---

Q5 (Multiple Choice, 2 points)

Question: In MuJoCo, what does data.qpos represent?

Options: a) Joint velocities b) Generalized positions (configuration) c) Applied torques d) Sensor readings

Correct Answer: b Explanation: qpos stores the generalized position coordinates—joint angles for revolute joints and positions for free-floating bodies. Learning Objective: Identify key components and capabilities of modern humanoid robot systems

---

Q6 (Short Answer, 4 points)

Question: Compare and contrast the challenges faced by Physical AI versus Software AI. Give a specific example for each.

Rubric:

• 2 points for correctly identifying differences
• 1 point for Physical AI example
• 1 point for Software AI example

Sample Answer: Physical AI must handle real-time constraints (can't pause to think longer), sensor noise, and physical consequences of failure. Software AI operates in deterministic digital environments where failures can be retried. Example Physical AI: A robot grasping an egg must control force precisely—too much breaks it, too little drops it. Example Software AI: A chess engine can take varying time per move and simply recalculate if interrupted. Learning Objective: Define Physical AI and distinguish it from traditional robotics and software AI

---

Q7 (Multiple Choice, 2 points)

Question: Which of the following is NOT a current application domain for humanoid robots?

Options: a) Manufacturing and logistics b) Healthcare and elder care c) Faster-than-light space travel d) Research and education

Correct Answer: c Explanation: While humanoids are used in manufacturing, healthcare, and research, faster-than-light travel is physically impossible and not a real application domain. Learning Objective: Identify key components and capabilities of modern humanoid robot systems

---

Q8 (Short Answer, 2 points)

Question: What is the simulation timestep in the MuJoCo humanoid model, and why does timestep matter for physics simulation?

Sample Answer: The default timestep is typically 0.002 seconds (2ms). Timestep matters because smaller timesteps provide more accurate physics simulation (especially for contact and fast dynamics) but require more computation. Learning Objective: Identify key components and capabilities of modern humanoid robot systems

---

Total Quiz Points: 20

Lab Assessment (35%)

Grading Rubric for Labs 01-01, 01-02, 01-03

Criterion	Weight	Excellent (90-100%)	Good (70-89%)	Satisfactory (50-69%)	Needs Work (<50%)
Code Functionality	40%	All code runs without errors, produces correct output	Minor errors, mostly correct	Some functions work	Major errors
Understanding	25%	Can explain code and concepts clearly	Explains most concepts	Partial understanding	Limited understanding
Documentation	15%	Well-commented, clear notebook narrat`ive	Adequate comments	Minimal documentation	No documentation
Checkpoints	20%	All checkpoints verified and documented	Most checkpoints complete	Some checkpoints done	Few checkpoints

Auto-Grading Tests

# tests/lab_01_01_tests.py
def test_mujoco_installed():
    import mujoco
```python
```python
    assert mujoco.__version__ >= "3.0.0"

def test_model_loads():
    import mujoco
    model = mujoco.MjModel.from_xml_path(
        mujoco.util.get_resource_path("humanoid/humanoid.xml")
    )
```python
```python
    assert model.nbody > 0

```python
```python
    assert model.njnt > 0

def test_simulation_runs():
    import mujoco
    model = mujoco.MjModel.from_xml_path(
        mujoco.util.get_resource_path("humanoid/humanoid.xml")
    )
    data = mujoco.MjData(model)
    initial_time = data.time
    mujoco.mj_step(model, data)
```python
```python
    assert data.time > initial_time

Simulation Project (35%)

Project: Humanoid Model Analysis

Description: Create a comprehensive analysis of the MuJoCo humanoid model, documenting its structure, simulating various initial conditions, and visualizing the results.

Deliverables

Deliverable	Format	Points
Model documentation	Markdown	20
Simulation script	Python	30
Results visualization	PNG/PDF plots	25
Written analysis	Markdown	25

Requirements

1. Model Documentation (20 points)

   • Document all bodies, joints, and actuators
   • Create a kinematic tree diagram
   • List all sensors available

2. Simulation Script (30 points)

   • Implement at least 3 different initial conditions
   • Run each for 5 seconds of simulated time
   • Log position, velocity, and energy data

3. Visualization (25 points)

   • Plot joint trajectories over time
   • Plot total energy (kinetic + potential)
   • Create comparison plots across initial conditions

4. Analysis (25 points)

   • Explain observed behaviors
   • Discuss energy conservation
   • Identify any unexpected results

Rubric

Criterion	Excellent	Good	Satisfactory	Needs Work
Completeness	All deliverables complete	Most complete	Partially complete	Missing items
Technical Accuracy	No errors	Minor errors	Some errors	Major errors
Insight	Deep analysis	Good observations	Basic analysis	Little analysis
Presentation	Professional quality	Good quality	Acceptable	Poor quality

Ethics Component (15%)

Format: Written Reflection with Peer Review

Length: 500-750 words Peer Reviews: 2 required

Prompt

Reflect on the ethical implications of pursuing Physical AI and humanoid robotics as a career or research field. Consider:

1. What responsibilities do roboticists have to society?
2. How should potential negative impacts (job displacement, safety risks) influence research directions?
3. What ethical frameworks or principles will guide your own work in this field?

Rubric

Criterion	Points	Description
Engagement with Material	30	References module content and case studies
Critical Thinking	30	Demonstrates nuanced analysis of trade-offs
Personal Reflection	20	Connects to own values and career goals
Writing Quality	10	Clear, well-organized writing
Peer Review Quality	10	Thoughtful feedback on others' reflections

Peer Review Guidelines

When reviewing peers' reflections:

• Identify one strength in their argument
• Ask one clarifying question
• Suggest one perspective they might not have considered

Export Formats

This assessment is available in:

• QTI 2.1 (IMS standard)
• Canvas quiz import
• Moodle GIFT format
• Blackboard import
• PDF (instructor version with answers)
• Gradescope (auto-grading enabled)

Alignment with Learning Objectives

Learning Objective	Quiz	Lab	Project	Ethics
Define Physical AI	Q1, Q6	✓	✓	✓
Historical evolution	Q3		✓
Key components	Q4, Q5, Q7, Q8	✓	✓
Role of embodiment	Q2	✓		✓