CAS Physical AI & Robotics

This program is designed for professionals who:

• Hold an engineering or science degree (e.g., electrical, mechanical, mechatronics, computer science, data science, physics, systems engineering) and may have little or only partial prior experience with AI or robotics.
• Work in one area relevant to Physical AI (hardware, software, or machine learning) and want to build the complementary skills needed to understand and implement complete robotic systems.
• Aim to transition into the growing field of robotics and embodied AI and require a structured, practice-oriented introduction that spans mechanical design, control, software architecture, and modern AI techniques.

By the end of the program, participants will be able to:

• Move beyond being users of robotic systems to become developers who can analyse, extend, and adapt robot capabilities to new tasks and environments regardless of whether they come from hardware, software, or machine learning backgrounds.
• Analyse robotic workspaces and tasks, select appropriate (mobile) robots, actuators, sensors, and hardware architectures, and design the corresponding control structures and real-time communication systems.
• Understand and compare different approaches to build intelligent robotic systems, from manipulators to mobile robots, and assess the trade-offs between classical and learning-based methods, as well as how to exploit the best of both worlds.
• Design and implement robotic software systems using the robot operating system (ROS), simulation, and distributed or cloud-enabled architectures, including the integration of external AI services.
• Build complete Physical AI systems (modern AI running on and controlling a robotic platform via ROS), from agentic approaches to end-to-end learning.
• Understand and apply modern learning-based methods in robotics, use reinforcement learning and foundation models for navigation and manipulation, fine-tune pretrained multimodal models, and evaluate the limitations as well as the opportunities and risks that generative AI brings into autonomous robotic systems.

Module “Introduction”

• Physical Design
• Computing and servers, Linux fundamentals, Python environment management, Docker
• Introduction to AI, practical supervised learning, neural networks, deep learning
• Robotics software fundamentals: ROS nodes and communication basics

Module “Robot Design, Actuators, Sensors and Control (HW)”

• Kinematic design of manipulators and mobile robots, coordinate transformations
• Actuators: motors, gears, drive systems
• Sensors: position, velocity, force, torque
• Motion control
• Hardware architecture and field buses for realtime communication

Module “Robot Software Systems and ROS (SW)”

• ROS Fundamentals and architecture
• Robot models, visualization, simulation
• Building ROS packages
• Robot pose, coordinate transformation systems
• SLAM, navigation, perception
• Software-level control
• Arm motion planning and execution (MoveIt, pick & place, Task Constructor)
• Deliberation
• Distributed Robotics

Module “Learning-Based Robotics and Embodied AI (ML)”

• Role of ML in perception–planning–control loops
• End-to-end learning in autonomous robots (incl. self-driving)
• Diffusion models and generative policies in robotic control, selfsupervised learning
• Reinforcement learning (RL) basics and sim-to-real transfer; model-free vs model-based RL; RL for manipulation tasks
• Embodied foundation models: vision-language-action models (VLAs) / Large behavior models (LBMs)
• Simulation tools for learning

The program combines lectures, case studies, and hands-on exercises with group work, self-study, and elements of e-learning. This blend ensures that participants can apply concepts directly to their professional environment.

Participants who complete the program successfully will be awarded the certificate «CAS ZHAW in Physical AI & Robotics». The program corresponds to 12 ECTS credits under the European Credit Transfer System.

Classes take place once a week on a Thursday. The program runs for five months. A detailed timetable is provided one month before the start. Breaks follow the official school holidays of the City of Winterthur.

The program is taught by a team of accomplished lecturers who combine strong academic credentials with extensive experience. Selected faculty members include:

• Prof. Dr. Marcel Honegger
• Dr. Jorge Pena Queralta
• Dr. Giovanni Toffetti Carughi
• Prof. Dr. Thilo Stadelmann

Applicants are expected to hold a university degree (University of Applied Sciences, University, ETH). Professionals with equivalent experience may also be admitted if they can demonstrate the required competencies. Basic programming skills and experience with Python are expected while exposure other programming languages language (C/C++) is beneficial. Affinity for mechatronics and data analysis, are recommended.