The Learning Autonomous Service Robots (LASR) team is a team of students competing in service robot challenges. LASR is the incarnation of the Sensible Robots research, and at the same time gives our undergraduate and master's students a chance to do real-world robot programming.
Dr Matteo Leonetti - Lecturer, team leader.
Dr Gerard Canal - post-doc, development leader.
Juan Camacho Mohedano - 3rd year UG student
Jared Swift - 3rd year UG student
The team will recruit the next group of students at KCL in January, after they completed the module "Introduction to Robotics".
The focus of our research is on adaptive decision making, to build systems that can plan quickly at different levels, improve over time, and constantly adapt to the people and the environment. Previous work (partly carried out by at the University of Texas at Austin), focuses on planning and learning algorithms for both symbolic and motion planning (see the research page, especially the section on service robots). In the near future, we intend to apply the same vision to human-robot interaction, merging an online and continuous learning component into planning for HRI.
With colleagues at UT Austin, we developed the actasp library for integrated reinforcement learning, reasoning, planning, and execution monitoring in Answer Set Programming (as part of this work). The library enables both the bwi_kr_execution and plan_execution packages in the BWI system, made publicly available by the group at UT. The code is platform independent and can be used on any robot, through implementing appropriate actions.
We contributed the learning extension of Petri Net Plans, maintained by the SPQR team at Sapienza.
We developed a package for object-aware navigation, which allows to tailor the behaviour to the type of obstacles. For instance, if a small object is blocking the path it can be pushed aside, while if it is a person the robot asks the person to kindly move out of the way.
Our winning team!