In recent years robots left controlled laboratory settings and started to address real world scenarios and situations. In order to tackle 'the wild', robots need to adapt their knowledge and their behaviour over time. Lifelong and online learning techniques are currently being established as one of the major research goals in mobile robotics. The rising interest in lifelong learning stems from the fact that all major mobile robot tasks, including mapping, navigation, manipulation, target tracking, planning or learning in general can be performed more accurately and more reliably when formulated and addressed as online learning problems. Given that modern mobile robot systems are being employed in highly dynamic environments such as cities, the capability of improving and refining the internal representation and performance of algorithms for mobile robots during their operation becomes a crucial requirement.
The aim of this workshop is to specifically show how the formulation as lifelong learning problems can be done, to give solutions and show results to these problems, and to create a platform for interchanging ideas between the different areas of mobile robotics on the topic of online and lifelong learning. Some of the leading researchers from robotics, computer vision and machine learning will give invited talks, but there will also be time for presentations of submitted and peer-reviewed contributions.
Invited talk preseantations online:
Francesco Orabona
Efficient and Principled Online Classification Algorithms for Lifelong Learning
Oliver Brock
Lifelong Learning in Autonomous Manipulation
Maxim Likachev
Lifelong Search-based Planning: From Incremental Planning to Planning with Experience