Curiosity driven reinforcement learning for motion planning on humanoids

Most previous work on artificial curiosity (AC) and intrinsic motivation focuses on basic concepts and theory. Experimental results are generally limited to toy scenarios, such as navigation in a simulated maze, or control of a simple mechanical system with one or two degrees of freedom. To study AC in a more realistic setting, we embody a curious agent in the complex iCub humanoid robot. Our novel reinforcement learning (RL) framework consists of a state-of-the-art, low-level, reactive control layer, which controls the iCub while respecting constraints, and a high-level curious agent, which explores the iCub's state-action space through information gain maximization, learning a world model from experience, controlling the actual iCub hardware in real-time. To the best of our knowledge, this is the first ever embodied, curious agent for real-time motion planning on a humanoid. We demonstrate that it can learn compact Markov models to represent large regions of the iCub's configuration space, and that the iCub explores intelligently, showing interest in its physical constraints as well as in objects it finds in its environment.

Speaking of supervision: A dialogic approach to building higher degree research supervision capacity in the creative arts

In the emergent field of creative practice higher degrees by research, first generation supervisors have developed new models of supervision for an unprecedented form of research that combines creative practice and written thesis. In a national research project, entitled 'Effective supervision of creative practice higher research degrees', we set out to capture and share early supervisors' insights, strategies and approaches to supporting their creative practice PhD students. From the insights we gained during the early interview process, we expanded our research methods in line with a distributed leadership model and developed a dialogic framework. This led us to unanticipated conclusions and unexpected recommendations. In this study, we primarily draw on philosopher and literary theorist Mikhail Bakhtin's dialogics to explain how giving precedence to the voices of supervisors not only facilitated the articulation of dispersed tacit knowledge, but also led to other 20 discoveries. These include the nature of supervisors' resistance to prescribed models, policies and central academic development programmes; the importance of polyvocality and responsive dialogue in enabling continued innovation in the field; the benefits to supervisors of reflecting, discussing and sharing practices with colleagues; and the value of distributed leadership and dialogue to academic development and supervision capacity building in research education.

Structures of three different neutral polysaccharides of Acinetobacter baumannii, NIPH190, NIPH201, and NIPH615, assigned to K30, K45, and K48 capsule types, respectively, based on capsule biosynthesis gene clusters

Neutral capsular polysaccharides (CPSs) were isolated from Acinetobacter baumannii NIPH190, NIPH201, and NIPH615. The CPSs were found to contain common monosaccharides only and to be branched with a side-chain 1→3-linked β-d-glucopyranose residue. Structures of the oligosaccharide repeat units (K units) of the CPSs were elucidated by 1D and 2D 1H and 13C NMR spectroscopy. Novel CPS biosynthesis gene clusters, designated KL30, KL45, and KL48, were found at the K locus in the genome sequences of NIPH190, NIPH201, and NIPH615, respectively. The genetic content of each gene cluster correlated with the structure of the CPS unit established, and therefore, the capsular types of the strains studied were designated as K30, K45, and K48, respectively. The initiating sugar of each K unit was predicted, and glycosyltransferases encoded by each gene cluster were assigned to the formation of the linkages between sugars in the corresponding K unit.