Skip to main content

Assessments

Grading Breakdown​

  • Weekly Labs (40%): 13 labs × 3% each (drop lowest 3 scores)
  • Module Projects (40%): 4 projects × 10% each
  • Capstone Project (20%): Final autonomous humanoid demonstration

Passing Grade: 70% overall

Project 1: ROS 2 Humanoid Control System​

Module: Module 1 - ROS 2 Due: End of Week 4 Weight: 10%

Objective​

Design a distributed ROS 2 system to control a simulated humanoid robot's joint movements using publishers, subscribers, services, and actions.

Deliverables​

  1. Architecture Diagram: ROS 2 node graph showing all nodes, topics, and services (PDF or PNG)
  2. Source Code: Python package with:
    • Joint state publisher node
    • Motion planning service
    • Emergency stop action server
    • Launch file for all nodes
  3. URDF Model: Humanoid robot description with 20+ degrees of freedom
  4. Demo Video: 2-minute screen recording showing:
    • Robot visualization in RViz2
    • Joint control via published commands
    • Service call for motion planning
    • Emergency stop action execution

Evaluation Criteria​

CriterionPointsDetails
Code Quality30PEP8 compliance, modularity, comments, type hints
Functionality40All requirements met, nodes communicate correctly
URDF Accuracy20Valid kinematics, proper joint limits, realistic inertia
Documentation10README with setup instructions, inline docstrings
Total100

Resources​

Project 2: Gazebo Digital Twin​

Module: Module 2 - Digital Twin Due: End of Week 7 Weight: 10%

Objective​

Create a high-fidelity physics simulation in Gazebo with virtual sensors and implement sensor fusion for autonomous obstacle avoidance.

Deliverables​

  1. Gazebo World File: SDF file with:
    • Realistic physics parameters (gravity, friction, timestep)
    • Environment with obstacles, stairs, or uneven terrain
    • Lighting and ground plane
  2. Sensor Integration:
    • LiDAR (360° scan, 10 Hz)
    • Depth camera (640x480, 30 Hz)
    • IMU (100 Hz)
  3. Sensor Fusion Node: Python node that combines sensor data for obstacle detection
  4. RViz2 Configuration: Save RViz config showing all sensor visualizations
  5. Demo Video: 3-minute demonstration of robot navigating environment using sensors

Evaluation Criteria​

CriterionPointsDetails
Physics Accuracy25Stable simulation, realistic contact dynamics
Sensor Configuration25All sensors publishing correct message types
Fusion Algorithm30Combines LiDAR + IMU for robust detection
Visualization10Clear RViz2 setup with all sensor data
Documentation10Instructions for running simulation
Total100

Resources​

Project 3: Isaac ROS VSLAM Integration​

Module: Module 3 - NVIDIA Isaac Due: End of Week 10 Weight: 10%

Objective​

Implement GPU-accelerated visual SLAM using Isaac ROS and deploy to NVIDIA Jetson hardware (or cloud instance).

Deliverables​

  1. VSLAM Setup:
    • Stereo camera configuration (RealSense D435i or simulated)
    • Isaac ROS VSLAM node running on Jetson Orin Nano
    • Parameter file tuned for environment
  2. Map Building: Generate map of test environment with loop closure
  3. Localization Demo: Show real-time pose estimation in RViz2
  4. Performance Report:
    • FPS (frames per second)
    • Latency (milliseconds)
    • Tracking accuracy (compare to ground truth)
  5. Demo Video: 2-minute video showing VSLAM in action

Evaluation Criteria​

CriterionPointsDetails
VSLAM Deployment30Successfully running on Jetson with correct topic connections
Map Quality25Loop closure detected, drift <5%
Performance20>20 FPS, <50ms latency
Documentation15Setup guide for replication
Analysis10Performance report with metrics
Total100

Resources​

Capstone: Autonomous Humanoid​

Module: Module 4 - VLA Due: End of Week 13 Weight: 20%

Objective​

Build an end-to-end voice-controlled humanoid robot that integrates all course modules: ROS 2, Gazebo simulation, Isaac perception, and VLA planning.

System Requirements​

Input: Voice command (e.g., "Robot, go to the kitchen and bring me the red cup")

Output: Autonomous execution with steps:

  1. Speech recognition (Whisper)
  2. Task planning (GPT-4 LLM)
  3. Navigation (Nav2 + VSLAM)
  4. Object detection (YOLOv8 or similar)
  5. Grasping (simplified or full MoveIt! integration)
  6. Return navigation
  7. Confirmation speech output

Deliverables​

  1. Source Code (GitHub repository):
    • All ROS 2 packages
    • Launch files
    • Configuration files (Nav2, sensors, VSLAM)
    • URDF/SDF models
    • README with setup instructions
  2. Documentation:
    • Architecture diagram (Mermaid or draw.io)
    • API documentation (function docstrings)
    • Troubleshooting guide
  3. Demo Video (2-3 minutes):
    • Show voice command input
    • Visualize LLM-generated plan
    • Time-lapse of robot execution
    • Code walkthrough (30 seconds)
  4. Evaluation Report:
    • Success metrics (completion rate, execution time)
    • Failure analysis
    • Future improvements

Test Scenarios (must pass 4/5)​

  1. Simple Fetch: "Bring me the red cup"
  2. Multi-Object: "Bring me the cup and the bottle"
  3. Recovery: Introduce obstacle mid-navigation, robot must replan
  4. Ambiguity: "Bring me something to drink" (robot asks for clarification)
  5. Failure Mode: Place object out of reach, verify graceful error handling

Evaluation Criteria​

CriterionPointsDetails
Functionality40System completes 4/5 test scenarios
Code Quality20Modularity, PEP8, comments, error handling
Integration15All modules (ROS 2, Gazebo, Isaac, VLA) working together
Documentation10Clear README, architecture diagram, API docs
Demo Video10Professional quality, shows end-to-end capability
Error Handling5Graceful recovery from failures (replanning, retries)
Total100

Resources​

Weekly Labs (13 Total)​

Lab 1: ROS 2 Publisher-Subscriber (Week 2)​

Create a temperature sensor publisher and a monitor subscriber that logs data to a file.

Submission: GitHub repo link

Lab 2: ROS 2 Service (Week 2)​

Implement a service for robot configuration (e.g., set max velocity).

Submission: Source code + demo video

Lab 3: Configurable Node with Parameters (Week 3)​

Build a node that reads parameters from YAML file and adjusts behavior at runtime.

Submission: Code + parameter file

Lab 4: Gazebo Custom World (Week 5)​

Create a Gazebo world with obstacles, stairs, and dynamic objects.

Submission: SDF file + screenshots

Lab 5: Sensor Integration (Week 6)​

Add LiDAR and depth camera to robot in Gazebo, visualize in RViz2.

Submission: Launch file + RViz config

Lab 6: Synthetic Data Generation (Week 8)​

Use Isaac Sim to generate 1000 labeled images with domain randomization.

Submission: Dataset (sample 100 images) + statistics

Lab 7: VSLAM Deployment (Week 9)​

Deploy Isaac ROS VSLAM and build map of simulated environment.

Submission: Saved map file + performance metrics

Lab 8: Nav2 Configuration (Week 10)​

Configure Nav2 costmaps and planners for humanoid navigation.

Submission: Parameter files + navigation demo video

Lab 9: Voice-Controlled Navigation (Week 11)​

Build system that navigates to locations via voice commands.

Submission: Source code + demo

Labs 10-13: Capstone Support Labs​

Incremental development of capstone project components (planning, perception, grasping, integration).

Submission: Progress reports + code milestones

Submission Guidelines​

  • Format: GitHub repository (public or private with instructor access)
  • Structure: Follow ROS 2 package conventions
  • Documentation: README.md must include:
    • Setup instructions
    • Dependencies
    • How to run
    • Expected output
  • Videos: Upload to YouTube/Vimeo, include link in README
  • Reports: PDF format, max 5 pages

Academic Integrity​

  • Collaboration is encouraged for learning, but submissions must be individual work
  • Cite all external resources (tutorials, Stack Overflow, etc.)
  • Use of AI assistants (ChatGPT, Copilot) is allowed but must be documented
  • Plagiarism will result in zero credit for the assignment

View weekly breakdown →

View hardware requirements →

Return to Homepage →