Search Unity

Become a Deep Reinforcement Learning expert

Master Deep Reinforcement Learning skills and implement state-of-the-art algorithms to build AI agents.

Learn more
In partnership with

Learn the most cutting-edge AI skills

From making it possible for a computer to beat the world’s best Go player to paving a path to Artificial General Intelligence, Deep Reinforcement Learning is one of the hottest fields in AI. And now there’s a hands-on way to develop this cutting-edge skill. In this program, you will learn how to apply:

  • Deep learning architectures to reinforcement learning tasks to build your own Deep Q-Network (DQN), which you can use to train an agent that learns intelligent behavior from raw sensory data.
  • Evolutionary algorithm and policy-gradient method theories, such as REINFORCE, DDPG, TRPO, and PPO, to research for your own algorithm, which you can use to train a simulated robotic agent to solve a complex task.
  • Reinforcement learning methods to applications that involve multiple, interacting agents, such as the coordination of autonomous vehicles.
  • Industry expertise from Unity and Udacity’s team of AI experts to develop professional deep reinforcement learning models.

Learn AI the project-based way

The best way to learn AI is to do AI. That’s why this program includes three in-depth projects that you can get expert feedback on and then add to your Github portfolio.


  • Intermediate to advanced Python experience. You are familiar with object-oriented programming. You can read and understand code written by others.
  • Intermediate statistics background. You are familiar with probability.
  • Intermediate knowledge of machine learning techniques. You can describe backpropagation, and have seen multiple examples of neural network architectures (like a CNN for image classification).
  • You have seen or worked with a deep learning framework like TensorFlow, Keras, or PyTorch.

Price: $999 USD

Time needed to complete all projects: Four months

  • Project 1: Navigation
  • Project 2: Continuous Control
  • Project 3: Collaboration and competition


Value-based methods

Master the foundations of reinforcement learning —from Markov Decision Processes to Bellman equations. Then, leverage Convolutional Neural Networks (CNNs) to train an agent that learns intelligent behaviors from sensory data.

What’s covered:

  • Monte Carlo Methods
  • Temporal-Difference Methods
  • Convolutional Neural Networks + PyTorch
  • Deep Q-Learning
Learn more

Continuous control

Policy-based methods

Learn the theory behind policy-based methods, such as evolutionary algorithms, stochastic policy search, and the REINFORCE algorithm. Then, use what you’ve learned to train a robotic arm to reach target locations, or train a four-legged creature to walk.

What’s covered:

  • Policy-based methods
  • Improving policy gradient methods
  • Actor-critic methods
Learn more

Collaboration and competition

Multi-agent reinforcement learning

Learn how to define a reinforcement learning task with multiple agents. Then, train a system of agents to demonstrate collaboration or cooperation on a complex task.

What's covered:

  • Multi-agent RL
  • Learning to collaborate
  • Learning to compete
Learn more

Explore other Learning and Certification products

Master Deep Reinforcement Learning

Learn the hottest field in AI, earn your Nanodegree credential, and build a portfolio of cutting-edge applications!

Learn more