• Fundamental methods for robot learning, learning in state spaces, learning with sensors and motors in reality.
• Learning in order to see: 3D understanding, localisation, image models with open vocabulary, image-language models.
• Learning in order to map: Neural world representations and mapping.
• Learning in order to act: Imitation learning and reinforcing learning with robots.
• Towards embodied AI: Large language models for robotics, foundation models for robotics and embodied image-language-action models.

After passing the course, the student shall be able to:

• explain the basic ideas and the challenges of learning for robots
• give an account of the concept embodied artificial intelligence and how robot learning can be combined with large-scale pre-trained neural networks such as multi-modal language models
• give an account of the theoretical background behind the methods for robot learning that are most common
• analyse advanced research in the area and critically evaluate the methods' weaknesses and strengths

in order to be able to

• implement, analyse and evaluate simple systems for robot learning
• absorb information about and read literature in the area
• implement methods based on new research and evaluate them.

• 12 in-person lectures (45min+45min) with 3 in-person quizes
• 6 in-person lab help sessions
• 3 labs (deliverable + oral examination)
• 9 in-person project help sessions
• 1 group project (deliverable + oral examination)

The course uses an interactive flipped class-room approach with assigned reading material consisting of slides, scientific papers and video tutorials. There is no text book since the field is moving so quickly. 

Students at KTH with a permanent disability can get support during studies from Funka:

Funka - compensatory support for students with disabilities

• LAB1 - Laboratory work, 4.5 credits, grading scale: P, F
• PRO1 - Project work, 3.0 credits, grading scale: A, B, C, D, E, FX, F

Based on recommendation from KTH’s coordinator for disabilities, the examiner will decide how to adapt an examination for students with documented disability.

The examiner may apply another examination format when re-examining individual students.

If the course is discontinued, students may request to be examined during the following two academic years.

To pass the course the students need at least E (50%) on the monitored quizes (assessing ILO1, ILO2), E on all threee labs (ILO3) and E on the project (ILO4) reflecting a basic understanding in each ILO. The latter two consist of deliverables with an oral comprehension assessment.

Higher course grades are an average of letter grades (A,C,E) given on:

• the three labs (assessing ILO3)
• project (ILO4)
• the three quiz grades (ILO1)

These are assessed as:

• Quiz grade C requires being able to explain most methods.
• Quiz grade A reqiures being able to explain nearly all methods.
• Lab grade C requires to have solved the intermediate difficulty lab exercises.
• Lab grade A requires to have solved the advanced difficulty lab exercises.
• Project grade C requires thorough evaluation of good quality (e.g. ablations).
• Project grade A requires to have implemented credible extension or improvement to existing method and evaluated the result (topic approved by teachers or selected from list of pre-approved topics).

• All members of a group are responsible for the group's work.
• In any assessment, every student shall honestly disclose any help received and sources used.
• In an oral assessment, every student shall be able to present and answer questions about the entire assignment and solution.

The course was given for the first time last year and the students gave it a high mark (4.3 of 5). This year we have attempted to adress some weaknesses identified - improving structure and scheduling of deadlines, as well as attempting to better homogenize difficulty across labs and projects. The lecture material has also been expanded to better cover some topics where students struggled to keep up last time: policy gradient methods, transformers and vision foundation models.