Skip to main content

Module 10: Capstone and career

This is where everything comes together. Across the earlier modules you learned each piece on its own: how robots move, how ROS 2 connects software, how to simulate, perceive, plan and learn, and how foundation models can turn language into action. In this module you combine those pieces into one working system, a voice-commanded robot assistant that hears a request, plans, drives, finds an object and picks it up.

The module also looks past the screen. A system that works in simulation is only half of the story. You will learn what changes on real hardware, how to think about safety and standards before a robot ever goes near people, and how to present your work so that others can see what you can do.

By the end of this module you can​

  • Specify, plan and build a complete robot system with measurable requirements.
  • Structure a multi-package ROS 2 workspace with clear interfaces and tests.
  • Evaluate a robot with logged trials and fix the biggest failure first.
  • Describe how to move from Gazebo to real hardware using ros2_control, safety stops and calibration.
  • Run a simple risk assessment and name the main robot-safety standards.
  • Present a portfolio and follow a 90-day plan toward your first robotics role.

Lessons​

LessonWhat you buildTime
Capstone briefThe project plan, requirements, milestones and an evaluation harness40 min
Capstone build guideThe workspace, interfaces and a validated mission orchestrator45 min
Deploying on hardwareA command watchdog and a staged hardware test plan45 min
Ethics, safety and standardsA risk register and a safety case summary40 min
Careers and portfolioA polished README, demo video and 90-day plan35 min

Module project​

The module project is the capstone itself: a voice-commanded robot assistant in simulation. A mobile manipulator in Gazebo Harmonic hears a spoken request such as "bring the red cup to the table", plans with an LLM, navigates with Nav2, perceives with a detector and picks and places with MoveIt 2.

You finish with a repository that builds and runs from one command, an evaluation of at least 20 logged trials with a failure breakdown, a short demo video and a written safety summary. Choose one stretch goal, such as a humanoid variant or a reduced run on real hardware, if you want to go further.

Stuck at any point? Ask Rivet, the course assistant, bottom-right.