Course Overview
This 13-week curriculum is structured to take you from zero robotics knowledge to building autonomous humanoid systems. Each week builds on previous concepts with a mix of theory, hands-on coding, and simulation exercises.
Curriculum Timeline
Weekly Breakdown
Weeks 1-2: Introduction to Physical AI
Foundation concepts for embodied intelligence
Week 1: Foundations
- What is Physical AI?
- Digital vs. embodied intelligence
- Key challenges in robotics
- History and evolution of humanoid robots
Week 2: Hardware & Sensors
- Robot anatomy: sensors, actuators, computation
- Sensor types: cameras, LiDAR, IMU, encoders
- Actuator types: DC motors, servos, steppers
- Computational platforms: Jetson, embedded systems
Learning Outcomes:
- Define Physical AI and its applications
- Identify core components of robot systems
- Understand sensor/actuator capabilities and limitations
- Recognize challenges unique to physical embodiment
Assessment: None (introductory material)
Module 1: ROS 2 Fundamentals (Weeks 3-5)
Master the Robot Operating System for production robotics
Week 3: ROS 2 Architecture
- Why ROS 2? Differences from ROS 1
- Distributed computing model
- Communication patterns: publish/subscribe, services, actions
- Quality of Service (QoS) policies
Week 4: Building ROS 2 Applications
- Creating nodes in Python and C++
- Publishing and subscribing to topics
- Implementing services and actions
- Parameter management and configuration
Week 5: Advanced ROS 2
- Custom message and service types
- Launch files for complex systems
- Lifecycle nodes and managed composition
- Debugging and introspection tools
Learning Outcomes:
- Design distributed robot applications using ROS 2
- Implement nodes, topics, services, and actions
- Create launch configurations for multi-node systems
- Debug ROS 2 applications using command-line tools
Assessment: Build a multi-node ROS 2 application for a simulated mobile robot with sensor fusion
Module 2: Robot Simulation (Weeks 6-7)
Test and iterate in high-fidelity simulation environments
Week 6: Gazebo Fundamentals
- Gazebo architecture and plugins
- Creating worlds and environments
- URDF and SDF robot modeling
- Integrating sensors (cameras, LiDAR) and physics
Week 7: Advanced Simulation
- Physics engine tuning (ODE, Bullet, DART)
- Sensor noise models
- Unity integration for visualization
- Sim-to-real gap considerations
Learning Outcomes:
- Create realistic simulation environments in Gazebo
- Model robots using URDF/SDF formats
- Integrate simulated sensors with ROS 2
- Understand and mitigate sim-to-real transfer challenges
Assessment: Build a complete Gazebo simulation of a mobile manipulator robot with cameras and LiDAR
Module 3: NVIDIA Isaac Platform (Weeks 8-10)
Leverage GPU acceleration for perception and learning
Week 8: Isaac Sim & SDK
- NVIDIA Omniverse and Isaac Sim overview
- Photorealistic rendering and sensor simulation
- Isaac SDK: building blocks for robotics
- ROS 2 integration with Isaac
Week 9: Perception & Manipulation
- GPU-accelerated computer vision pipelines
- Object detection and pose estimation
- Grasp planning and execution
- SLAM and navigation
Week 10: Reinforcement Learning
- RL fundamentals for robotics
- Isaac Gym: parallel RL training
- Training locomotion policies
- Sim-to-real transfer techniques
Learning Outcomes:
- Set up and use NVIDIA Isaac Sim
- Implement GPU-accelerated perception pipelines
- Train RL policies for robot tasks
- Transfer learned behaviors from simulation to reality
Assessment: Train a reinforcement learning policy for robot manipulation and deploy in Isaac Sim
Module 4: Humanoid Development (Weeks 11-13)
Build complete autonomous humanoid systems
Week 11: Kinematics & Dynamics
- Forward and inverse kinematics
- Jacobians and velocity control
- Dynamics: mass, inertia, forces
- Whole-body controllers
Week 12: Bipedal Locomotion
- Zero Moment Point (ZMP) stability
- Gait generation and foot placement
- Balance and disturbance rejection
- Stair climbing and uneven terrain
Week 13: Conversational Robotics
- Large language model integration (GPT-4, LLaMA)
- Speech recognition (Whisper) and synthesis (TTS)
- Multimodal interaction: vision + language
- Context-aware conversation management
Learning Outcomes:
- Implement kinematics and dynamics for humanoid robots
- Develop stable bipedal locomotion controllers
- Integrate conversational AI for natural interaction
- Coordinate full-body motion with manipulation and locomotion
Assessment: Final Capstone Project (see below)
Capstone Project: Autonomous Humanoid Robot
Objective: Build a complete autonomous humanoid system that can:
- Navigate to a target location using SLAM
- Detect and grasp objects using vision-based manipulation
- Respond to voice commands in natural language
- Maintain balance during locomotion and manipulation
Requirements:
- Simulation: Must work in Isaac Sim or Gazebo
- Perception: Object detection + SLAM
- Manipulation: Pick and place at least 3 objects
- Locomotion: Walk at least 5 meters, handle obstacles
- Conversation: Respond to 10 predefined voice commands
Deliverables:
- Source code (ROS 2 workspace)
- Simulation world and robot model (URDF/USD)
- Demo video (3-5 minutes)
- Technical report (5-10 pages)
Evaluation Criteria (100 points):
- Functionality (40 pts): Does it work reliably?
- Code Quality (20 pts): Clean, modular, well-documented
- Technical Depth (20 pts): Advanced techniques used appropriately
- Creativity (10 pts): Novel approaches or features
- Documentation (10 pts): Clear explanation of design decisions
Time Commitment
Per Week:
- Lectures/Reading: 3 hours
- Hands-on Labs: 4 hours
- Assignments/Projects: 3 hours
- Total: ~10 hours/week
Total Course: ~130 hours + ~40 hours capstone = 170 hours
Prerequisites Recap
Required:
- Python programming (intermediate level)
- Linear algebra (vectors, matrices, transformations)
- Basic physics (Newton's laws, kinematics)
- Command-line comfort (bash, ssh)
Recommended (but not required):
- C++ programming
- Calculus (derivatives, optimization)
- Control theory (PID controllers)
- Machine learning basics
Software Setup (before Week 1):
- Ubuntu 22.04 (native or WSL2)
- ROS 2 Humble
- Gazebo Garden
- Python 3.10+
- VS Code or PyCharm
See the Software Architecture chapter for detailed installation guides.
Learning Pathways
Different learners can follow different paths through the material:
Path 1: Beginner (All Weeks)
Complete all 13 weeks sequentially. Ideal for students with no robotics experience.
Path 2: Accelerated (10 Weeks)
Skip Weeks 1-2 (Physical AI Intro) if you already understand robotics fundamentals. Focus on ROS 2, simulation, and Isaac platform.
Path 3: Specialization Tracks
Software Engineering Focus:
- Weeks 3-5 (ROS 2) + Week 6-7 (Simulation) + Capstone
AI/ML Focus:
- Weeks 1-2 (Physical AI) + Weeks 8-10 (Isaac/RL) + Week 13 (Conversational AI) + Capstone
Mechanical/Controls Focus:
- Weeks 6-7 (Simulation) + Weeks 11-12 (Kinematics/Locomotion) + Capstone
What Makes This Curriculum Unique?
- End-to-End: From zero to autonomous humanoid—no gaps
- Simulation-First: Test everything in simulation before touching hardware
- Modern Stack: ROS 2 (not ROS 1), Isaac Sim (not Gazebo-only), GPT integration
- Hands-On: Every concept has a runnable code example
- Scalable: Works with zero hardware or full robot lab
- Conversational AI: Integrate LLMs for natural human-robot interaction
- GPU-Accelerated: Leverage NVIDIA Isaac for 10-100x speedups
Next Steps
👉 Learning Outcomes — Detailed skills you'll master
👉 Why Physical AI? — The future of embodied intelligence
👉 Week 1-2: Physical AI Intro — Start the journey!