AI in Robotics

Pantelis Monogioudis, Ph.D

Course overview

A comprehensive journey in Robotics through the lens of AI.

9 modules
43 lessons
10 hr 29 min
Start the course
Module 1

Course Introduction

Robotic agents that can perceive and act in the world using AI.

  1. The Robotic AI Agent

    A practical map for navigating the illusion that you actually have any agency at all.

    14:36
  2. Development Environment Setup

    Setting up the ROS2 Docker-based development environment.

    14:36
  3. Mathematical Prerequisites

    What you need to know before diving into the course material.

    14:36
Module 2

Foundations: Statistical Learning Theory

From Vapnik to Hinton, the mathematical underpinnings of modern AI.

  1. The learning problem

    The Vapnik block diagram.

    14:52
  2. Linear Regression

    Extracting non-linear patterns with linear models.

    14:52
  3. Gradient Descent

    Optimizing complicated functions with iterative methods.

    14:52
  4. Entropy

    Informarion theory principles.

    14:52
  5. Maximum Likelihood Estimation

    The workhorse of statistical modeling.

    14:52
  6. Binary Classification

    Binary classification and Logistic Regression.

    14:52
Module 3

Neural Networks

How neural networks are trained to extract the right features.

  1. Feature extraction

    Using a simple network to understand how features are extracted.

    14:36
  2. Multiclass Classifier

    A simple multiclass classifier example.

    14:36
  3. Backpropagation

    How to calculate gradients in a neural network.

    14:36
  4. Regularization

    How to regulate the complexity of a neural network.

    14:36
Module 4

Convolutional Neural Networks

Extracting spatial features with convolutional layers.

  1. How we understand scenes ?

    Human Perception and Imaging.

    14:36
  2. Convolution and Correlation

    A linear operation for extracting spatial features.

    14:36
  3. CNN Architectures

    Looking inside a CNN layyer and understanding architectural patterns.

    14:36
  4. Image Classification

    Image classification with data augmentation.

    14:36
  5. What CNNs Learn

    Visualizing the features learned by CNNs.

    14:36
  6. ResNets

    Residual Networks and skip connections.

    14:36
Module 5

Object Detection

Localizing objects in images.

  1. Introduction to Object Detection.

    Object detection in a physical security application.

    14:36
  2. Computer vision datasets

    What types of annotations are used in computer vision ?

    14:36
  3. Region-based Object Detectors

    R-CNN, Fast R-CNN, Faster R-CNN.

    14:36
Module 6

Recursive State Estimation

Presenting Hidden Markov Models and Kalman filters as a tool for state estimation.

  1. Introduction to State Estimation with HMM.

    Introducing Hidden Markov Models.

    14:36
  2. The Bayes Filter

    Implementing the Bayes filter algorithm.

    14:36
  3. An example of Discrete Bayes Filter

    Discrete Bayes localization notebook .

    14:36
  4. Continuous state space and the Kalman filter

    Localizing a drone under Gaussian assumptions.

    14:36
  5. A Kalman filter example

    Kalman localization notebook.

    14:36
Module 7

Global Planning

Using domain-specific languages and search algorithms for global planning.

  1. Introduction to Planning.

    Typical planning problems and PDDL.

    14:36
  2. Planning Domain Definition Language

    The constructs of PDDL

    14:36
  3. Forward Search Algorithms

    Finding global planning solutions with forward search.

    14:36
  4. The A* Algorithm

    Using heuristics to guide forward search.

    14:36
Module 8

Multimodal Reasoning and Transformers

Using transformer architectures for multimodal reasoning.

  1. Introduction to Transformers for language modeling.

    The transformer architecture and the simple attention mechanism.

    14:36
  2. The Learnable Attention Mechanism.

    Implementing the scaled dot-product self attention mechanism.

    14:36
  3. Multi-Head Self Attention

    Using multiple attention heads to capture different aspects of input sequences.

    14:36
  4. ViT Transformers and CLIP - Part 1

    ViT Transformers and contrastive learning with CLIP.

    14:36
Module 9

Markov Decision Processes

Markov Decision Processes as a framework for sequential decision making under uncertainty.

  1. Introduction to MDPs - Part 1

    Defining Markov Decision Processes.

    14:36
  2. Introduction to MDPs - Part 2

    Defining Markov Decision Processes.

    14:36
  3. The Bellman Expectation Equations - Part 1

    Deriving the Bellman Expectation Equations.

    14:36
  4. The Bellman Expectation Equations - Part 2

    Deriving the Bellman Expectation Equations.

    14:36
  5. Policy Evaluation - Part 1

    Using the Bellman Expectation Equations for Policy Evaluation.

    14:36
  6. Policy Evaluation - Part 2

    Using the Bellman Expectation Equations for Policy Evaluation.

    14:36
  7. Bellman Optimal Value Functions

    Deriving the Bellman Optimality Equations.

    14:36
  8. Policy Iteration and Value Iteration

    Using the Bellman Optimality Equations for optimal control.

    14:36