AI Robotics Course

Lesson 1 • Core AI Concepts for Robotics

Introduces machine learning, reasoning, and perception as applied to robots. Connects AI theory directly to robotic behaviour and decision-making.

Lesson 2 • Ethics and Safety in AI Robotics

Examines responsibility, bias, and safety standards in autonomous systems. Establishes an ethical lens applied consistently across all chapters.

Lesson 3 • Types of Robots and Use Cases

Surveys mobile, manipulator, collaborative, and soft robots. Frames domain-specific applications students will encounter throughout the course.

Lesson 4 • History and Evolution of Robotics

Traces robotics from mechanical automata to autonomous systems. Provides timeline context that anchors all subsequent technical concepts.

Lesson 1 • Calculus and Optimization

Reviews derivatives, gradients, and optimisation methods used in learning algorithms. Prepares students for gradient-based training in neural network chapters.

Lesson 2 • Geometry and Spatial Reasoning

Covers coordinate frames, rotations, and homogeneous coordinates for 3D space. Essential for robot pose representation and path planning.

Lesson 3 • Linear Algebra Essentials

Covers vectors, matrices, and transformations central to robot kinematics. Directly supports coordinate frame calculations in later chapters.

Lesson 1 • Forward Kinematics

Derives end-effector position from joint parameters using Denavit-Hartenberg notation. Establishes the computational foundation for all motion planning work.

Lesson 2 • Inverse Kinematics

Solves joint configurations from desired end-effector poses using analytical and numerical methods. Directly enables goal-directed robot arm control.

Lesson 3 • Robot Dynamics and Equations of Motion

Applies Newton-Euler and Lagrangian formulations to model forces and torques. Provides the physics basis for controller design in the next section.

Lesson 4 • Trajectory Planning and Interpolation

Generates smooth joint and Cartesian trajectories using polynomial and spline methods. Connects kinematics to executable motion commands.

Lesson 1 • Deep Learning for Visual Perception

Trains convolutional neural networks for object detection, segmentation, and depth estimation. Connects vision models to downstream robot decision-making.

Lesson 2 • Sensor Fusion Techniques

Combines multi-modal sensor streams using Kalman and particle filters. Produces robust state estimates that feed planning and control modules.

Lesson 3 • 3D Perception and Point Cloud Processing

Processes LiDAR and RGB-D point clouds for scene understanding and object localisation. Enables robots to reason about 3D structure in unstructured environments.

Lesson 4 • Sensor Modalities and Data Acquisition

Surveys LiDAR, cameras, IMUs, and tactile sensors with their noise characteristics. Establishes data quality awareness critical for all perception algorithms.

Lesson 1 • Graph-Based SLAM

Formulates SLAM as a pose graph optimisation problem and solves it with nonlinear least squares. Enables globally consistent large-scale mapping.

Lesson 2 • Map Management and Long-Term Navigation

Addresses dynamic environments, map updates, and multi-session mapping. Prepares students for real-world deployment where environments change over time.

Lesson 3 • Occupancy Grid Mapping

Builds probabilistic 2D and 3D environment maps from sensor data. Provides the map representation used in path planning chapters.

Lesson 4 • Probabilistic Localization

Uses Bayesian filters to estimate robot pose from sensor observations. Grounds all subsequent mapping work in rigorous state estimation.

Lesson 1 • Potential Fields and Reactive Navigation

Uses attractive and repulsive potential fields for real-time obstacle avoidance. Complements global planners with fast local reactive behaviours.

Lesson 2 • Trajectory Optimization

Optimises trajectories for smoothness, time, and energy using CHOMP and TrajOpt. Produces dynamically feasible paths ready for controller execution.

Lesson 3 • Search-Based Planning Algorithms

Implements A*, Dijkstra, and D* Lite on discrete graphs and grids. Establishes algorithmic planning intuition before sampling-based methods.

Lesson 4 • Sampling-Based Motion Planning

Applies RRT, RRT*, and PRM to high-dimensional configuration spaces. Enables planning for robot arms and mobile platforms in cluttered scenes.

Lesson 1 • Classical PID Control

Designs proportional-integral-derivative controllers for position and velocity loops. Provides the control baseline extended by all advanced methods in this chapter.

Lesson 2 • Force and Impedance Control

Regulates contact forces and compliant behaviour for manipulation and human interaction. Essential for safe operation in contact-rich tasks.

Lesson 3 • Robust and Adaptive Control

Handles model uncertainty and disturbances with sliding mode and adaptive schemes. Ensures stable performance when robot parameters vary or are unknown.

Lesson 4 • Whole-Body and Multi-Robot Control

Coordinates full-body motion for legged robots and task allocation across robot teams. Scales control concepts to complex, multi-degree-of-freedom systems.

Lesson 1 • Safe and Sample-Efficient RL

Applies constrained RL, model-based RL, and meta-learning to improve safety and efficiency. Prepares students for real-hardware deployment with limited trial budgets.

Lesson 2 • Simulation Environments for RL

Configures physics simulators and robotic environments for scalable RL training. Addresses the sim-to-real gap that limits deployment of trained policies.

Lesson 3 • Deep RL Algorithms

Implements DQN, PPO, SAC, and TD3 for continuous robotic control tasks. Provides algorithm selection criteria based on action space and sample efficiency.

Lesson 4 • Imitation and Inverse RL

Learns policies from expert demonstrations using behavioural cloning and inverse RL. Reduces sample complexity for complex manipulation tasks.