Trailer-tractor vehicles are the backbone of logistical operations, delivering goods to their destination. Despite their importance, publicly available research in mechanisms for making such vehicles autonomous is sparse. Furthermore, navigating through roundabouts is one of the most challenging aspects of driving such vehicles, due to the necessary lane changes, trajectories and tight curvatures, together with dynamic aspects such as coordinating with other traffic. In this work, we propose a reinforcement learning based framework to train a tractor-trailer vehicle to navigate through roundabout intersections. We provide the first tractor-trailer vehicle model for the high-fidelity CAR Learning to Act (CARLA) simulator together with the first dataset of roundabout scenarios. A model was trained using the Proximal Policy Optimisation algorithm, obtaining a testing success rate of 77% while maintaining minimal lateral deviation, mimicking human-like behaviour.

Placement is the initial step of Printed Circuit Board (PCB) physical design and demands considerable time and domain expertise. Placement quality impacts the performance of subsequent tasks, and the generation of an optimal placement is known to be, at the very least, NP-complete. While stochastic optimisation and analytic techniques have had some success, they often lack the intuitive understanding of human engineers. In this study, we propose a novel end-to-end Machine Learning (ML) approach to learn fundamental placement techniques and use experience to optimise PCB layouts efficiently. To achieve this, we formulate the PCB placement problem as a Markov Decision Process (MDP) and use Reinforcement Learning (RL) to learn general placement techniques. The agent-driven data collection process generates highly diverse and consistent data points sufficient for learning general policies without expert knowledge under the guidance of an adaptive reward signal. Compared to state-of-the-art sim-ulated annealing approaches on unseen circuits, the resulting policies trained with TD3 and SAC, on average, yield 17% and 21 % reduction in post-routing wirelength. Qualitative analysis shows that the policies learn fundamental placement techniques and demonstrate an understanding of the underlying problem dynamics. Collectively, they demonstrate emergent collaborative or competitive behaviours and faster placement convergence, sometimes exceeding an order of magnitude.

Music recommendation systems typically use collaborative filtering to determine which songs to recommend to their users. This mechanism matches a user with listeners that have similar tastes, and uses their listening history to find songs that the user will probably like. The fundamental issue with this approach is that artists already need to have a significant user following to get a fair chance of being recommended. This is known as the music cold-start problem. In this work, we investigate the possibility of making music recommendations based on audio content so that new artists still get a good chance of being recommended, even if they do not have a sufficient number of listeners yet.

Required concurrency can cause actions to interfere with running continuous effects. This interference can modify the rate of change, including the polarity, of a continuous effect. In this work, we propose a mechanism to support discrete interference of rates of change caused by instantaneous actions, the start and end endpoints of other durative actions, and numeric timed initial fluents. Current temporal planners have very limited support for such numeric dynamics. COLIN reduces a temporal numeric planning problem to a linear program (LP), but operates on an implicit assumption that the rate of change of a durative action’s continuous effect is constant throughout its execution. In this work we propose some enhancements to the algorithms used in COLIN, in order to support discrete interference of continuous effects, and a new planner, DICE, was developed to implement them.