Kathryn Kasmarik

Sherif Abdelfattah, Kathryn Kasmarik, Jiankun Hu

Multi-objective Markov decision processes are a special kind of multi-objective optimization problem that involves sequential decision making while satisfying the Markov property of stochastic processes. Multi-objective reinforcement learning methods address this problem by fusing the reinforcement learning paradigm with multi-objective optimization techniques. One major drawback of these methods is the lack of adaptability to non-stationary dynamics in the environment. This is because they adopt optimization procedures that assume stationarity to evolve a coverage set of policies that can solve the problem. This paper introduces a developmental optimization approach that can evolve the policy coverage set while exploring the preference space over the defined objectives in an online manner. We propose a novel multi-objective reinforcement learning algorithm that can robustly evolve a convex coverage set of policies in an online manner in non-stationary environments. We compare the proposed algorithm with two state-of-the-art multi-objective reinforcement learning algorithms in stationary and non-stationary environments. Results showed that the proposed algorithm significantly outperforms the existing algorithms in non-stationary environments while achieving comparable results in stationary environments.

Duy Tung Nguyen, Kathryn Kasmarik, Hussein Abbass

Neural-based learning agents make decisions using internal artificial neural networks. In certain situations, it becomes pertinent that this knowledge is re-interpreted in a friendly form to both the human and the machine. These situations include: when agents are required to communicate the knowledge they learn to each other in a transparent way in the presence of an external human observer, in human-machine teaming settings where humans and machines need to collaborate on a task, or where there is a requirement to verify the knowledge exchanged between the agents. We propose an interpretable knowledge fusion framework suited for neural-based learning agents, and propose a Priority on Weak State Areas (PoWSA) retraining technique. We first test the proposed framework on a synthetic binary classification task before evaluating it on a shepherding-based multi-agent swarm guidance task. Results demonstrate that the proposed framework increases the success rate on the swarm-guidance environment by 11% and better stability in return for a modest increase in computational cost of 14.5% to achieve interpretability. Moreover, the framework presents the knowledge learnt by an agent in a human-friendly representation, leading to a better descriptive visual representation of an agent's knowledge.

Vu Phi Tran, Matthew A. Garratt, Kathryn Kasmarik et al.

This paper proposes a novel swarm-based control algorithm for exploration and coverage of unknown environments, while maintaining a formation that permits short-range communication. The algorithm combines two elements: swarm rules for maintaining a close-knit formation and frontier search for driving exploration and coverage. Inspired by natural systems in which large numbers of simple agents (e.g., schooling fish, flocking birds, swarming insects) perform complicated collective behaviors for efficiency and safety, the first element uses three simple rules to maintain a swarm formation. The second element provides a means to select promising regions to explore (and cover) by minimising a cost function involving robots' relative distance to frontier cells and the frontier's size. We tested the performance of our approach on heterogeneous and homogeneous groups of mobile robots in different environments. We measure both coverage performance and swarm formation statistics as indicators of the robots' ability to explore effectively while maintaining a formation conducive to short-range communication. Through a series of comparison experiments, we demonstrate that our proposed strategy has superior performance to recently presented map coverage methodologies and conventional swarming methods.

Tung D. Nguyen, Kathryn E. Kasmarik, Hussein A. Abbass

It is well-known that information loss can occur in the classic and simple Q-learning algorithm. Entropy-based policy search methods were introduced to replace Q-learning and to design algorithms that are more robust against information loss. We conjecture that the reduction in performance during prolonged training sessions of Q-learning is caused by a loss of information, which is non-transparent when only examining the cumulative reward without changing the Q-learning algorithm itself. We introduce Differential Entropy of Q-tables (DE-QT) as an external information loss detector to the Q-learning algorithm. The behaviour of DE-QT over training episodes is analyzed to find an appropriate stopping criterion during training. The results reveal that DE-QT can detect the most appropriate stopping point, where a balance between a high success rate and a high efficiency is met for classic Q-Learning algorithm.

Hung The Nguyen, Tung Duy Nguyen, Vu Phi Tran et al.

The control and guidance of multi-robots (swarm) is a non-trivial problem due to the complexity inherent in the coupled interaction among the group. Whether the swarm is cooperative or non-cooperative, lessons can be learnt from sheepdogs herding sheep. Biomimicry of shepherding offers computational methods for swarm control with the potential to generalize and scale in different environments. However, learning to shepherd is complex due to the large search space that a machine learner is faced with. We present a deep hierarchical reinforcement learning approach for shepherding, whereby an unmanned aerial vehicle (UAV) learns to act as an aerial sheepdog to control and guide a swarm of unmanned ground vehicles (UGVs). The approach extends our previous work on machine education to decompose the search space into a hierarchically organized curriculum. Each lesson in the curriculum is learnt by a deep reinforcement learning model. The hierarchy is formed by fusing the outputs of the model. The approach is demonstrated first in a high-fidelity robotic-operating-system (ROS)-based simulation environment, then with physical UGVs and a UAV in an in-door testing facility. We investigate the ability of the method to generalize as the models move from simulation to the real-world and as the models move from one scale to another.

Duy T. Nguyen, Kathryn E. Kasmarik, Hussein A. Abbass

On the one hand, artificial neural networks (ANNs) are commonly labelled as black-boxes, lacking interpretability; an issue that hinders human understanding of ANNs' behaviors. A need exists to generate a meaningful sequential logic of the ANN for interpreting a production process of a specific output. On the other hand, decision trees exhibit better interpretability and expressive power due to their representation language and the existence of efficient algorithms to transform the trees into rules. However, growing a decision tree based on the available data could produce larger than necessary trees or trees that do not generalise well. In this paper, we introduce two novel multivariate decision tree (MDT) algorithms for rule extraction from ANNs: an Exact-Convertible Decision Tree (EC-DT) and an Extended C-Net algorithm. They both transform a neural network with Rectified Linear Unit activation functions into a representative tree, which can further be used to extract multivariate rules for reasoning. While the EC-DT translates an ANN in a layer-wise manner to represent exactly the decision boundaries implicitly learned by the hidden layers of the network, the Extended C-Net combines the decompositional approach from EC-DT with a C5 tree learning algorithm to form decision rules. The results suggest that while EC-DT is superior in preserving the structure and the fidelity of ANN, Extended C-Net generates the most compact and highly effective trees from ANN. Both proposed MDT algorithms generate rules including combinations of multiple attributes for precise interpretations for decision-making.

Hung Nguyen, Vu Tran, Tung Nguyen et al.

In apprenticeship learning (AL), agents learn by watching or acquiring human demonstrations on some tasks of interest. However, the lack of human demonstrations in novel tasks where they may not be a human expert yet, or when it is too expensive and/or time consuming to acquire human demonstrations motivated a new algorithm: Apprenticeship bootstrapping (ABS). The basic idea is to learn from demonstrations on sub-tasks then autonomously bootstrap a model on the main, more complex, task. The original ABS used inverse reinforcement learning (ABS-IRL). However, the approach is not suitable for continuous action spaces. In this paper, we propose ABS via Deep learning (ABS-DL). It is first validated in a simulation environment on an aerial and ground coordination scenario, where an Unmanned Aerial Vehicle (UAV) is required to maintain three Unmanned Ground Vehicles (UGVs) within a field of view of the UAV 's camera (FoV). Moving a machine learning algorithm from a simulation environment to an actual physical platform is challenging because `mistakes' made by the algorithm while learning could lead to the damage of the platform. We then take this extra step to test the algorithm in a physical environment. We propose a safety-net as a protection layer to ensure that the autonomy of the algorithm in learning does not compromise the safety of the platform. The tests of ABS-DL in the real environment can guarantee a damage-free, collision avoidance behaviour of autonomous bodies. The results show that performance of the proposed approach is comparable to that of a human, and competitive to the traditional approach using expert demonstrations performed on the composite task. The proposed safety-net approach demonstrates its advantages when it enables the UAV to operate more safely under the control of the ABS-DL algorithm.