SST · BCSAI · Course Structure

Robotics & Automation.

The complete, syllabus-driven structure of the course — meta, objectives, methodology, assessment, and all 30 sessions grouped by module, with a key-concepts glossary and an annotated bibliography.
Looking for the playground? Jump to the interactive view.

00 · Overview

What this course is

A robot is a computer with the physical ability to interact with the world — a goal-oriented machine that can sense, plan, and act. It senses its environment through sensors, fuses that information with a goal to plan an action, then drives actuators to carry it out, and repeats the loop continuously. This course covers every concept needed to make a robot do exactly that in simulated environments. Theory-based lectures are complemented by simulation exercises in ROS (Robot Operating System), the de facto standard software for robotics research. It pairs with the experimental Robotics & Automation LAB (OACT.M.A_C1_448152), where the same concepts are put into practice on real-robot hardware.

The structure mirrors that sense → plan → act loop. After a Linux and ROS foundation, the spine of the term walks the full robot software stack: describe the robot (URDF), place it in space (TF and kinematics), act on the world (control), move through it (navigation), and sense it (vision). A mid-term checkpoint then gives way to a research-frontiers seminar series and a capstone paper analysis — moving students from operators to critical readers of robotics research.

Program
BCSAI — Computer Science & AI
Course code
RA-CSAI.4.M.A
Intake
SEP-2025
Academic year
2025–26
Area
Computer Science & AI
Sessions
30
Credits
6.0 ECTS
Workload
~150 hours
Degree course
Fourth year
Semester
2nd
Category
Compulsory
Language
English
Prof. Eduardo Castelló Ferrer
ecastello@faculty.ie.edu · office hours on request

B.Sc. (Hons.) in intelligent systems (University of Portsmouth, 2007); M.Eng. and Ph.D. in robotics engineering (Osaka University, 2011 & 2016). His interests span robotics, cryptography, and complex systems. A former Marie Curie Fellow at the MIT Media Lab — where he pioneered combining distributed robotic systems with blockchain — he is now an assistant professor at IE and a research fellow with the MIT Connection Science group, working on security, behavior, and business models for robotics via novel cryptographic methods.

01 · Learning Objectives

By the end of the course you can…

02 · Methodology & Assessment

How learning and grading work

IE University's method is collaborative, active, and applied: students build knowledge and sharpen skills through lectures, discussions, in-class and asynchronous exercises, field work, group work, and individual study — roughly 150 hours total. The professor's role is to lead and guide students toward the learning objectives rather than simply transmit content. Grades come from intermediate tests, the final group presentation, and class participation.

Learning activities — weighting & hours
Lectures · 20h
13.3%
Discussions · 10h
6.7%
In-class exercises · 45h
30.0%
Async / field work · 30h
20.0%
Group + individual · 45h
30.0%
Final grade — weighting
Intermediate tests
50%
Group presentation
40%
Class participation
10%

Intermediate tests

50% of final grade

Deliverable: a session-19 mid-term, run as an individual ROS programming exercise, plus lab exercises completed individually with results presented.

Evaluated on: demonstrated proficiency with the ROS framework — correct use of nodes, topics/services/actions, and the ability to make a simulated robot sense, plan, and act.

Group presentation

40% of final grade

Deliverable: in groups of ~5, a deep scientific SWOT analysis of a robotics research paper chosen from candidates surfaced across sessions 20–29; presented in sessions 29–30.

Evaluated on: depth of analysis — clearly arguing the paper's strengths, weaknesses, opportunities, and threats. GenAI may not be used for the presentation.

Class participation

10% of final grade

Deliverable: presence in class plus contributions to discussion throughout the term.

Evaluated on: quality, not quantity — frequent speakers are not automatically rated higher. Students are encouraged to contribute from the very first sessions.

03 · Program

All 30 sessions, by module

The course runs as five arcs: setup & foundations, a long core ROS spine (description → kinematics → control → navigation → vision), a mid-term checkpoint, a research-frontiers seminar series, and a keynote & capstone finale. Each module opens with a short overview and a set of learning outcomes; each session lists its objective, the concepts it covers, the core technical idea where one applies, a key idea or concrete example, and annotated readings. Where a session maps onto an interactive demo, follow the blue tag to poke at it.

Module A

Setup & Foundations

Sessions 1–2

Orientation and the Linux groundwork everything else stands on. Before a robot can sense, plan, or act, you need to be at home on the operating system and command line it actually runs on.

Learning outcomes
  • Explain what robotics is and place the course's topics within the sense–plan–act framing.
  • Navigate a Linux filesystem, manage file permissions, and read/write simple bash scripts.
  • Connect to a robot's onboard computer remotely over ssh.
Session 01 · live in-person

Course kick-off & the robotics field

Objective: Frame the field and set expectations for the term.

  • What robotics is — the history (Karel Čapek's 1920 play R.U.R. coined "robot") and the modern definition of a goal-oriented machine that senses, plans, and acts.
  • Instructor's background — robotics, cryptography, and complex systems, framing the research flavour of the course.
  • Admin — class schedule, software installation, and opening the online accounts needed for the simulation work.
  • Roadmap — an overview of every topic the course will cover.
Key idea A robot is not its body but its control loop: the same sense→plan→act cycle scales from a vacuum to a Mars rover. The whole course is one long elaboration of that loop.
Session 02 · live in-person

GNU/Linux essentials

Objective: Be fluent on the Linux command line a robot runs on.

  • Filesystem — navigating and interacting with the Linux directory tree (ROS itself is built on Linux conventions).
  • Permissions — the read/write/execute model that governs which processes may touch which files and devices.
  • Bash scripting — writing simple programs to automate setup and launch steps.
  • Remote access — connecting to a robot's onboard computer over ssh, since real robots are rarely controlled at a keyboard.
  • A wrap-up programming example tying the concepts together.
Key idea Almost every robot you will ever deploy is "headless" — you reach it over the network with ssh and drive it with scripts, not a mouse. Command-line fluency is a robotics skill.
Module B

Core ROS — Sense, Plan, Act

Sessions 3–18

The spine of the course: from ROS plumbing and robot description through kinematics, control, navigation, and vision — then a review. Each block adds one layer to a robot that can ultimately perceive its world and move purposefully through it.

Learning outcomes
  • Build ROS systems using nodes communicating over topics, services, and actions.
  • Describe a robot's links and joints in URDF and reason about its degrees of freedom.
  • Use the TF tree to track coordinate frames over time and solve forward/inverse kinematics for an arm.
  • Design and tune PID control loops and apply them on the Tiago / Pepper platforms.
  • Build a map, localize, and plan a path for a mobile base (Robotnik SUMMIT-XL).
  • Apply OpenCV for color, edge, object, and face detection in a robotics context.
Sessions 03–04 · live in-person

Introduction to ROS

Objective: Understand what ROS is, why we use it, and its core mechanics.

  • What is ROS — a middleware framework that connects many small programs (nodes) into one robot, so teams can reuse drivers and algorithms instead of rewriting them.
  • Topics — named asynchronous message streams (publish/subscribe) for continuous data like sensor readings.
  • Services — synchronous request/response calls for one-off queries.
  • Actions — long-running, preemptible goals with continuous feedback (e.g. "navigate to X").
  • A closing programming example summarizing the concepts.
Core concept · ROS graph A running ROS system is a graph of nodes exchanging typed messages. Choose the right channel: topics for streams, services for quick queries, actions for goals that take time and may be cancelled.
Quigley et al. — Programming Robots with ROS demo: nodes, topics, messages →

Reading: Quigley et al. — chapters on the ROS graph, rostopic, and writing publisher/subscriber nodes map directly onto these two sessions.

Sessions 05–06 · live in-person

URDF — describing a robot

Objective: Define a robot's structure and physical properties in an XML-like format.

  • Visual model — building the robot's geometry from links (rigid bodies) and joints (the connections between them).
  • Physical properties — adding collision geometry, inertia, and friction so the model behaves correctly in simulation.
  • URDF in ROS — feeding the description to the rest of the ecosystem; a simulated wrap-up example.
Core concept · links & joints URDF (Unified Robot Description Format) is a tree of links joined by joints. Each non-fixed joint adds a degree of freedom (DoF) — the number of independent ways the robot can move — which is exactly what kinematics and control will later reason about.
Siciliano & Khatib — Springer Handbook demo: degrees of freedom →

Reading: Siciliano & Khatib — the kinematics/modeling chapters underpin how a link-and-joint description translates into a movable mechanism.

Sessions 07–08 · live in-person

TF (Transforms & Frames) + kinematics

Objective: Track multiple coordinate frames over time and reason about arm kinematics.

  • Coordinate frames — expressing where things are relative to each other (e.g. an object in the camera frame vs. the base frame).
  • Robot & joint states — reading the configuration of every joint to know the robot's current pose.
  • Kinematic models — the math linking joint angles to end-effector pose; a simulated arm manipulation example.
Core concept · TF tree & FK/IK The TF tree stores time-stamped transforms between frames so any pose can be expressed in any other frame. Forward kinematics (FK) maps joint angles → end-effector pose; inverse kinematics (IK) solves the reverse — which joint angles reach a target pose — and may have several or no solutions.

Reading: Lynch & Park — the forward- and inverse-kinematics chapters are the canonical reference for the FK/IK math demonstrated here.

Sessions 09–10 · live in-person

ROS control — making robots act

Objective: Build control loops and apply them to a simulated manipulation task.

  • Control loops — continuously comparing a desired setpoint with a measured state and correcting the error.
  • PID dynamics — how the proportional, integral, and derivative terms shape response speed, steady-state error, and overshoot.
  • Custom controllers — applied on the Tiago / Pepper platforms in a simulated manipulation task.
Core concept · PID control A PID controller drives error e(t) to zero with u = Kp·e + Ki·∫e dt + Kd·de/dt: P reacts to present error, I cancels accumulated steady-state error, D damps overshoot by anticipating change. Tuning the three gains is the central skill.
Lynch & Park — Modern Robotics demo: PID control loop →

Reading: Lynch & Park — the robot-control chapters cover feedback control and the PID gains tuned in the demo.

Sessions 11–12 · live in-person

ROS navigation — mapping & planning (I)

Objective: Make a robot explore and navigate an environment.

  • Mapping — building a map of a controlled environment from sensor data.
  • Localization — estimating where the robot is within that map.
  • Path planning — computing a route from a start to a desired destination; applied on the Robotnik SUMMIT-XL in simulation.
Core concept · SLAM & planning SLAM (Simultaneous Localization And Mapping) builds a map while tracking the robot's pose within it — solving the chicken-and-egg problem of needing a map to localize and a pose to map. A path planner (e.g. A*, Dijkstra) then searches that map for a collision-free route, balancing path cost against search effort.
Siegwart et al. — Autonomous Mobile Robots demo: A* path planning → demo: differential drive →

Reading: Siegwart et al. — the localization, mapping, and planning chapters are the primary reference for this navigation block.

Sessions 13–14 · live in-person

ROS navigation — mapping & planning (II)

Objective: Deepen the navigation stack and consolidate planning skills.

  • Further map building, localization, and path planning, hardening the stack against harder environments.
  • Extended simulated navigation on the Robotnik SUMMIT-XL, including re-planning when the world changes.
Core concept · global vs. local planning Real navigation splits into a global planner (a route over the known map) and a local planner (short-horizon obstacle avoidance that adjusts to dynamic obstacles). When the map changes, the robot must re-plan rather than blindly follow the old route.
Siegwart et al. — Autonomous Mobile Robots demo: A* re-planning →

Reading: Siegwart et al. — continue with the planning and obstacle-avoidance chapters to support the deeper navigation exercises.

Sessions 15–16 · live in-person

ROS + OpenCV — robot vision

Objective: Let a robot see and recognize its environment.

  • Low-level vision — recognizing patterns, colors, and borders (edges) in an image.
  • Detection — finding objects, people, and faces in a robotics context.
  • A simulated wrap-up example connecting a camera feed to a robot's behavior.
Core concept · color spaces & edges Filtering by color is far more robust in HSV (hue–saturation–value) than RGB, because hue separates "what color" from "how bright". Edge detection finds boundaries where image intensity changes sharply — the raw material for recognizing objects and shapes.
Siciliano & Khatib — Springer Handbook demo: HSV color filtering →

Reading: Siciliano & Khatib — the sensing and machine-vision chapters cover the perception pipeline behind these OpenCV exercises.

Sessions 17–18 · live in-person

ROS review & consolidation

Objective: Revisit all 15 theory sessions and prep for the mid-term.

  • Concept review — a pass over every key concept from ROS plumbing through vision.
  • Practical exercises — focused on control and data acquisition across the different robot platforms.
  • A simulated wrap-up example integrating the full stack.
Key idea By now the pieces form one pipeline: describe (URDF) → place (TF/kinematics) → act (control) → move (navigation) → see (vision). The mid-term tests whether you can assemble them into a working node.
Module C

Mid-Term Checkpoint

Session 19

A practical demonstration of ROS proficiency — the half-way gate between the hands-on core and the research-oriented second half of the term.

Learning outcomes
  • Independently write a ROS program that makes a simulated robot sense, plan, and act.
  • Demonstrate command of nodes, topics/services/actions, and at least one of kinematics, control, navigation, or vision under exam conditions.
Session 19 · live in-person

Mid-term exam

Objective: Demonstrate proficiency with the ROS framework via an individual programming exercise. Counts toward the 50% intermediate-tests weighting.

Format: an individual, hands-on programming exercise — not a written test. You build and run code rather than describe it.

Key idea The exam rewards working code over recall: the goal is to prove you can turn the sense–plan–act concepts into a running ROS node on your own.
Module D

Research Frontiers

Sessions 20–25

A seminar series on state-of-the-art robotics research. Each session surveys one emergent field; paper candidates for the final group presentation are surfaced throughout, so students start reading critically while the topics are fresh.

Learning outcomes
  • Survey the state of the art across several emergent robotics research fields.
  • Read a robotics paper critically and identify a candidate for the capstone SWOT analysis.
  • Connect frontier research back to the sense–plan–act fundamentals of the core module.
Session 20 · live in-person

New directions in robotics research

Objective: Survey emergent research fields and the latest advances opening up the discipline.

  • An entry point to the seminar series: a map of where active robotics research is heading.
  • How to read a research paper — and what makes one a good candidate for the capstone analysis.
Key idea This session opens the "paper candidate" pipeline that runs to session 29 — start a shortlist now.
Session 21 · live in-person

Industry 4.0 & the future of work

Objective: Explore how AI, IoT, and automation reshape work-related processes.

  • Industry 4.0 — the convergence of AI, the Internet of Things, and automation in connected factories.
  • Research where robots improve work-related processes, and what that means for the future of work.
Context · Industry 4.0 "Industry 4.0" names the fourth industrial revolution: cyber-physical systems where networked sensors, robots, and AI coordinate production with minimal human intervention.
Session 22 · live in-person

Human augmentation & biomechanics

Objective: Study where the human body meets the artificial to enhance its capabilities.

  • Biomechanics — modeling the body's movement to design devices that work with it.
  • Research where exoskeletons, prosthetics, and wearables augment human capability.
Context · augmentation Human augmentation extends the body rather than replacing it — an exoskeleton, for example, amplifies strength while leaving the human in control of intent.
Session 23 · live in-person

Bio-inspired robotics

Objective: Examine robot design and function inspired by biological systems.

  • How animal morphology and behavior inform locomotion, grasping, and sensing.
  • Research where biology is a design template — from insect gaits to soft, compliant bodies.
Context · bio-inspiration Bio-inspired design borrows principles evolution already optimized — e.g. compliant (soft) actuation for safe, adaptable contact — rather than copying nature literally.
Session 24 · live in-person

Cognitive robotics

Objective: Look at robots that achieve intelligent behavior, learning, and reasoning in complex environments.

  • How learning and reasoning let a robot handle situations its programmers did not foresee.
  • Research where perception, memory, and decision-making combine into adaptive behavior.
Context · cognitive robotics Cognitive robotics moves beyond fixed control laws toward systems that learn and reason — closing the sense–plan–act loop with adaptation rather than pre-scripted plans.
Session 25 · live in-person

Human–robot interaction & collaborative robotics

Objective: Study robots that share space with humans and interact with them directly.

  • HRI — designing communication, safety, and trust for robots that work near or with people.
  • Cobots — collaborative robots built to share a workspace safely with human co-workers.
Context · cobots A cobot is engineered for shared space: force-limited motion and rich sensing let it stop or yield on contact, so safety is built into the hardware, not just the software.
Module E

Keynotes & Capstone

Sessions 26–30

Industry perspectives, then the final group research-paper analysis that ties the term together.

Learning outcomes
  • Relate course concepts to real-world industry practice through expert keynotes.
  • Deliver a structured SWOT analysis of a robotics research paper as a team.
  • Defend the analysis clearly to an audience — the 40% capstone deliverable.
Session 26 · live in-person

Industry keynote I

Objective: Hear an industry expert connect course concepts to robotics practice in the field.

Session 27 · live in-person

Industry keynote II

Objective: A second industry keynote, broadening the view of where robotics is applied commercially.

Session 28 · live in-person

Industry keynote III

Objective: A third industry keynote, closing out the guest-speaker series before the capstone.

Sessions 29–30 · live in-person

Final group presentation

Objective: In groups of ~5, deliver a deep scientific SWOT analysis (strengths, weaknesses, opportunities, threats) of a chosen robotics research paper. Worth 40% of the final grade.

  • Paper choice — selected from candidates provided across sessions 20–29.
  • Analysis — a structured argument about the paper's contribution and limitations, not a summary.
  • Constraint — GenAI may not be used for the presentation or group submission.
Method · SWOT A SWOT read frames a paper along four axes: internal Strengths and Weaknesses (what the work does and fails to do) and external Opportunities and Threats (what it could enable and what could undermine it).
04 · Key Concepts

Glossary

A quick reference for the technical vocabulary used across the program — the terms a robotics paper, a ROS tutorial, or a lab demo will assume you already know.

Sense–plan–act loop
The core robotics cycle: perceive the world with sensors, fuse that with a goal to plan an action, execute it with actuators, repeat. The whole course elaborates this loop.
ROS middleware
The Robot Operating System — a framework that wires many small programs (nodes) into one robot via standard messaging, enabling code reuse across projects.
Node
A single ROS process responsible for one job (a driver, a planner, a controller). A robot is a graph of cooperating nodes.
Topic
A named, asynchronous message stream using publish/subscribe — ideal for continuous data such as sensor readings or velocity commands.
Service
A synchronous request/response call between nodes — used for one-off queries that return immediately.
Action
A long-running, preemptible goal with continuous feedback (e.g. "navigate to a pose"), which can be monitored and cancelled mid-execution.
URDF
Unified Robot Description Format — an XML-like language describing a robot as links (rigid bodies) connected by joints, plus collision and inertial properties.
Degrees of freedom DoF
The number of independent parameters needed to specify a robot's configuration — typically one per movable joint.
TF tree
ROS's transform system: a time-stamped tree of coordinate frames that lets any pose be re-expressed in any other frame as the robot moves.
Forward kinematics FK
Computing the end-effector's pose from known joint angles — a direct, single-valued mapping.
Inverse kinematics IK
The reverse problem: finding joint angles that place the end-effector at a target pose. May have multiple, one, or no solutions.
Workspace
The set of all poses a robot's end-effector can physically reach given its links and joint limits.
Control loop
A feedback process that repeatedly compares a desired setpoint with the measured state and issues corrections to drive the error to zero.
PID controller
A controller combining Proportional, Integral, and Derivative terms: u = Kp·e + Ki·∫e + Kd·de/dt — tuned to balance speed, steady-state error, and overshoot.
SLAM
Simultaneous Localization And Mapping — building a map of an unknown environment while tracking the robot's pose within it.
Localization
Estimating the robot's pose (position and orientation) within a known map from sensor data.
Path planning
Searching a map for a collision-free route from a start to a goal; A* and Dijkstra are classic graph-search planners.
Global vs. local planner
The global planner finds a route over the known map; the local planner handles short-horizon obstacle avoidance and dynamic obstacles, triggering re-planning when needed.
Differential drive
A mobile-base design steered by varying the speeds of two independently driven wheels — turning by making one wheel spin faster than the other.
OpenCV
The open-source computer-vision library used in ROS for color, edge, object, and face detection.
HSV color space
Hue–Saturation–Value — separates color identity (hue) from brightness, making color-based detection more robust than in RGB.
Edge detection
Finding boundaries in an image where intensity changes sharply — a building block for shape and object recognition.
Cobot
A collaborative robot designed to share a workspace safely with humans, using force-limited motion and sensing to react to contact.
HRI
Human–Robot Interaction — the study of how robots communicate, cooperate, and build trust with people.
Industry 4.0
The fourth industrial revolution: networked cyber-physical systems combining AI, IoT, and automation across connected production.
SWOT analysis
A four-axis critique — Strengths, Weaknesses, Opportunities, Threats — applied to a research paper in the capstone presentation.
05 · Bibliography

Recommended reading — annotated

Four recommended texts, each anchoring a different stretch of the course. The note on each entry says what it covers and which sessions it supports.