Knowledge representation, reasoning, and learning for human-robot collaboration

Robots assisting humans in complex domains receive different descriptions of uncertainty and incomplete domain knowledge. This includes commonsense knowledge about the domain, especially default knowledge that holds true in all but a few exceptional situations, e.g., statements such as "textbooks are usually in the library but cookbooks are in the kitchen". The robot also extracts quantitative descriptions of knowledge by processing sensor inputs, e.g., statements such as "I am 90% certain I saw the AI book on the table". The existing knowledge may need to be revised over time, and humans may not have the expertise or time to provide elaborate and accurate feedback. In addition, the robot receives far more raw data from sensors and humans than it can process efficiently. Widespread use of robots thus poses fundamental knowledge representation, reasoning, and learning challenges that can be mapped to the following research questions:

(*) How to best enable robots to represent and reason reliably and efficiently with qualitative and quantitative descriptions of knowledge and uncertainty?

(*) How to best enable robots to learn interactively from multimodal sensor inputs and limited feedback from human participants who lack domain expertise?

The PhD student involved in this research project will develop algorithms to address these questions in specific application domains, e.g., robots collaborating with humans in offices, homes, or disaster rescue sites. The student will implement and evaluate these algorithms on different mobile robot platforms (e.g., wheeled, humanoid) collaborating with humans to describe scenes, and find and manipulate domain objects.

Funding Notes

The position offered is for four years of part time (75%) study with 456 teaching hours per year. The value of the award is £18,552 pa.

2:1 Honours undergraduate degree and/or postgraduate degree with Distinction (or an international equivalent).

Proficiency in probability theory, statistics, calculus, and linear algebra, and excellent object-oriented programming expertise. Prior knowledge of logics, machine learning, robotics, or AI will be a plus but is not essential.

If your first language is not English and you have not studied in an English-speaking country, you will have to provide an English language qualification.

References

Mohan Sridharan and Ben Meadows. Knowledge Representation and Interactive Learning of Domain Knowledge for Human-Robot Interaction. In Advances in Cognitive Systems Journal, 7:1-20, November 2018.

Mohan Sridharan, Michael Gelfond, Shiqi Zhang, and Jeremy Wyatt. REBA: A Refinement-Based Knowledge Representation and Reasoning Architecture for Robots. Technical report on arXiv, September 2018.

Mohan Sridharan, Ben Meadows and Rocio Gomez. What can I not do? Towards an Architecture for Reasoning about and Learning Affordances. In the International Conference on Automated Planning and Scheduling, Pittsburgh, USA, June 18-23, 2017.

Shiqi Zhang, Mohan Sridharan, and Jeremy Wyatt. Mixed Logical Inference and Probabilistic Planning for Robots in Unreliable Worlds. In the IEEE Transactions on Robotics (T-RO), 31(3):699-713, June 2015.