Edge AI & Autonomous Navigation

NVIDIA Jetson with CUDA

5 sessions
6 hours each
30 Hrs total
Live Online

Full price

$ 1700

Bundle and save

4-times payment

$ 425
Klarna

Buy now, pay later

$ 141
Affirm

Upcoming Enrollment

November 30th - December 4th

5 sessions 6 hours each

Online, US Eastern Time

Share it

Build an autonomous ground robot that maps unknown spaces, spots hazards with on-board GPU vision, and steers around moving obstacles.

Learn how to push your robot to 30+ FPS on NVIDIA Jetson with CUDA, TensorRT, and Python, and prove it with real benchmarks. Go past tutorials and build real autonomy with ROS 2, SLAM, Nav2, and edge AI running on a Pi 5 and Jetson Orin. You'll finish with a containerized, CI-tested robot and results you can measure and put on your resume.

Edge AI & Autonomous Navigation takes you from a pile of hardware to a robot that drives itself. You'll network a multi-board ROS 2 system, build maps with SLAM, write A* from scratch, and tune Nav2 to dodge moving obstacles. You'll also speed up vision on NVIDIA Jetson with CUDA and INT8 TensorRT. Along the way you'll port hot paths to Python, add failsafes, and ship everything in Docker with GitHub Actions CI. You'll leave with a working robot, real performance numbers, and skills hiring managers ask about.

Skills You Will Walk Away With

  1. 1

    Distributed Robotics Systems / Networking

    • Configuring a multi-machine ROS 2 ARM Pi 5 + NVIDIA Jetson
    • DDS discovery
    • DDS Quality-of-Service tuning
    • Diagnosing which QoS mismatch
    • Static IP assignment on a robot LAN
    • Multi-host process orchestration
    • Headless operations of embedded devices
    • NTP/chrony
    • Camera streams
  2. 2

    Computer Vision

    • OpenCV in Python and API via the port work
    • Camera capture pipelines and V4L2 device handling
    • Camera calibration
    • Color-space conversion
    • Image preprocessing for inference
    • Object detection inference
    • RViz2 image panels
  3. 3

    GPU Acceleration & Edge AI

    • CUDA programming model — kernels, threads, blocks, grids, warps
    • GPU memory hierarchy
    • CuPy for GPU-accelerated array computation (drop-in NumPy replacement)
    • Numba JIT compilation and custom kernels
    • Host↔device memory transfer management
    • TensorRT engine building and ONNX model
    • Post-training quantization
    • Jetson power modes
    • NVIDIA Jetson Orin platform and JetPack SDK
    • Thermal and power-envelope management
  4. 4

    Sensor Fusion

    • Time-synchronizing heterogeneous sensor streams
    • Handling variable-rate sensors
    • TF2 transform trees applied to fusion
    • Designing a perception
    • Combining vision detections with range/odometry data into a single actionable state
    • Reasoning about sensor confidence and disagreement
  5. 5

    SLAM & Mapping

    • SLAM Toolbox
    • Occupancy grid representation
    • Costmap generation
    • Saving, serializing, and reloading maps
    • Localization against a known map (AMCL)
    • Loop closure
    • Odometry drift measurement
    • Diagnosing a map
  6. 6

    Path Planning & Autonomous Navigation

    • Dijkstra and A* algorithms
    • Heuristic design
    • Grid-based planning
    • Nav2 stack architecture
    • Behavior Trees
    • Recovery behaviors
    • Costmap2D layer configuration
    • Local controller tuning (DWB / MPPI)
    • Dynamic obstacle avoidance
    • ROS 2 action

Schedule & Enrollment

Instructors

Art Yudin

Questions? Check here