SQUIRREL is an EU project that is now in its final year. A robot is deployed in nurseries in Europe with a focus on having the robot -- named Kenny -- interact with children by playing games, regulating their emotions, and become a peer with the children.
Our focus in the first couple of years has been on the task-related issues: tidying the room; uncluttering heaps; looking for unknown objects in a known environment; learning new properties of objects, etc. The last year we switched our focus to the children by planning to regulate their emotions using the PAD model and have the planner generate behaviours to create relationships with children like reciprocity.
The main outputs created so far are:
- A Factorised Contingency Planner.
- Planning framework that can perform emotional regulation based on the PAD model.
- Recommender system that allows the planner to learn missing information from the domain model.
- Method to find novel objects in a known environment.
More information can be found on: http://www.squirrel-project.eu/
PANDORA was an EU project with the aim of creating a framework that provides persitent autonomy for AUVs. The mission was to maintain a subsea facility: inspecting structures to check there are no leaks; operate maintenance panels by turning valves; deal with marine sea life; and manage energy by recharging the AUV when needed.
The project was a success with our main outputs:
- A visualiser that was build using the Dreaded Portal Engine that allowed us to demonstrate multi-day missions.
- A tactical / strategic framework plan for multi-day missions.
- Opportunistic strategy to exploit "low probability high reward" events.
- ROSPlan, a ROS module that enables planning inside ROS projects.
More information can be found on: http://persistentautonomy.com/
Autonomous Systems Underpinning Research (ASUR) is a Defence Science and Technology Laboratory programme which aims to develop an underpinning science and technology base to facilitate the research, development, production, deployment and operational use of intelligent unmanned systems by UK Armed Forces. In this context we are working on the capabilities of an UAV to search the seabed for safe passages for ships.
Lifted planning
For my PhD work I worked on lifted planning and created a planner called L-RPG that can solve larger planning problems that the state-of-the-art. It exploits near symmetry relationships; It can detect when two objects are similar enough that they can be treated as being symmetrical.
Created two lifted heuristics:
- Lifted RPG (based on the FF heuristic)
- Lifted CG (based on the Fast-Downward heuristic)
It has been published at a few conferences and a journal paper is in the works. My PhD thesis can be found here: https://kclpure.kcl.ac.uk/portal/en/theses/lifted-heuristics(27d32f95-9859-4ca7-b7ca-b89e4d1fb977).html