The number of people with dementia (PwD) residing in nursing homes (NH) increases rapidly. Behavioural disturbances (BDs) such as wandering and aggressions are the main reasons to hospitalise these people. Social robots could help to resolve these BDs by performing simple interactions with the patients. This paper examines whether self-learning algorithms could be designed to select the robotic interactions, preferred by the patients, during an intervention. K-armed bandit algorithms were compared in simulated environments for single and multiple patients to find the beneficial learning agents and action selection policies. The single patient tests show the advantages of selecting actions according to an Upper Confidence Bound (UCB) policy, while the multi-patient tests analyse the benefits of using additional, contextual information. Afterwards, the learning application was provided with a framework to operate in more realistic situations. We expect that this framework can be used for personalised interactions in many different healthcare domains.

Scalable and effective exploration remains a key challenge in reinforcement learning (RL). While there are methods with optimality guarantees in the setting of discrete state and action spaces, these methods cannot be applied in high-dimensional deep RL scenarios. As such, most contemporary RL relies on simple heuristics such as epsilon-greedy exploration or adding Gaussian noise to the controls. This paper introduces Variational Information Maximizing Exploration (VIME), an exploration strategy based on maximization of information gain about the agent's belief of environment dynamics. We propose a practical implementation, using variational inference in Bayesian neural networks which efficiently handles continuous state and action spaces. VIME modifies the MDP reward function, and can be applied with several different underlying RL algorithms. We demonstrate that VIME achieves significantly better performance compared to heuristic exploration methods across a variety of continuous control tasks and algorithms, including tasks with very sparse rewards.

A key challenge in the realization of autonomous vehicles is the machine's ability to perceive its surrounding environment. This task is tackled through a model that partitions vehicle camera input into distinct semantic classes, by taking into account visual contextual cues. The use of structured machine learning models is investigated, which not only allow for complex input, but also arbitrarily structured output. Towards this goal, an outdoor road scene dataset is constructed with accompanying fine-grained image labelings. For coherent segmentation, a structured predictor is modeled to encode label distributions conditioned on the input images. After optimizing this model through max-margin learning, based on an ontological loss function, efficient classification is realized via graph cuts inference using alpha-expansion. Both quantitative and qualitative analyses demonstrate that by taking into account contextual relations between pixel segmentation regions within a second-degree neighborhood, spurious label assignments are filtered out, leading to highly accurate semantic segmentations for outdoor scenes.

Existing robot guides are able to offer a tour of a building, such as a museum, bank, science center, to a single person or to a group of participants. Usually the tours are predefined and there is no support for dynamic interactions between multiple robots. This paper focuses on distributed collaboration between several robot guides providing a building tour to groups of participants. Semantic techniques are adopted in order to formally define the tour topics, available content on a specific topic, and the robot and human profiles including their interests and content knowledge. The robot guides select different topics depending on their participants' interests and prior knowledge. Optimization of the topics of interests is achieved through exchange of participants between the robot guides whenever in each others neighborhood. Evaluation of the implemented algorithms presents a 90% content coverage of relevant topics for the individual participants.