Ivana Dusparic

Babatunji Omoniwa, Boris Galkin, Ivana Dusparic

In this letter, we study the energy efficiency (EE) optimisation of unmanned aerial vehicles (UAVs) providing wireless coverage to static and mobile ground users. Recent multi-agent reinforcement learning approaches optimise the system's EE using a 2D trajectory design, neglecting interference from nearby UAV cells. We aim to maximise the system's EE by jointly optimising each UAV's 3D trajectory, number of connected users, and the energy consumed, while accounting for interference. Thus, we propose a cooperative Multi-Agent Decentralised Double Deep Q-Network (MAD-DDQN) approach. Our approach outperforms existing baselines in terms of EE by as much as 55 -- 80%.

Jasmina Gajcin, Ivana Dusparic

While AI algorithms have shown remarkable success in various fields, their lack of transparency hinders their application to real-life tasks. Although explanations targeted at non-experts are necessary for user trust and human-AI collaboration, the majority of explanation methods for AI are focused on developers and expert users. Counterfactual explanations are local explanations that offer users advice on what can be changed in the input for the output of the black-box model to change. Counterfactuals are user-friendly and provide actionable advice for achieving the desired output from the AI system. While extensively researched in supervised learning, there are few methods applying them to reinforcement learning (RL). In this work, we explore the reasons for the underrepresentation of a powerful explanation method in RL. We start by reviewing the current work in counterfactual explanations in supervised learning. Additionally, we explore the differences between counterfactual explanations in supervised learning and RL and identify the main challenges that prevent the adoption of methods from supervised in reinforcement learning. Finally, we redefine counterfactuals for RL and propose research directions for implementing counterfactuals in RL.

Helio N. Cunha Neto, Ivana Dusparic, Diogo M. F. Mattos et al.

Fast identification of new network attack patterns is crucial for improving network security. Nevertheless, identifying an ongoing attack in a heterogeneous network is a non-trivial task. Federated learning emerges as a solution to collaborative training for an Intrusion Detection System (IDS). The federated learning-based IDS trains a global model using local machine learning models provided by federated participants without sharing local data. However, optimization challenges are intrinsic to federated learning. This paper proposes the Federated Simulated Annealing (FedSA) metaheuristic to select the hyperparameters and a subset of participants for each aggregation round in federated learning. FedSA optimizes hyperparameters linked to the global model convergence. The proposal reduces aggregation rounds and speeds up convergence. Thus, FedSA accelerates learning extraction from local models, requiring fewer IDS updates. The proposal assessment shows that the FedSA global model converges in less than ten communication rounds. The proposal requires up to 50% fewer aggregation rounds to achieve approximately 97% accuracy in attack detection than the conventional aggregation approach.

Jasmina Gajcin, James McCarthy, Rahul Nair et al.

A well-defined reward function is crucial for successful training of an reinforcement learning (RL) agent. However, defining a suitable reward function is a notoriously challenging task, especially in complex, multi-objective environments. Developers often have to resort to starting with an initial, potentially misspecified reward function, and iteratively adjusting its parameters, based on observed learned behavior. In this work, we aim to automate this process by proposing ITERS, an iterative reward shaping approach using human feedback for mitigating the effects of a misspecified reward function. Our approach allows the user to provide trajectory-level feedback on agent's behavior during training, which can be integrated as a reward shaping signal in the following training iteration. We also allow the user to provide explanations of their feedback, which are used to augment the feedback and reduce user effort and feedback frequency. We evaluate ITERS in three environments and show that it can successfully correct misspecified reward functions.

Jasmina Gajcin, Ivana Dusparic

While reinforcement learning (RL) algorithms have been successfully applied to numerous tasks, their reliance on neural networks makes their behavior difficult to understand and trust. Counterfactual explanations are human-friendly explanations that offer users actionable advice on how to alter the model inputs to achieve the desired output from a black-box system. However, current approaches to generating counterfactuals in RL ignore the stochastic and sequential nature of RL tasks and can produce counterfactuals that are difficult to obtain or do not deliver the desired outcome. In this work, we propose RACCER, the first RL-specific approach to generating counterfactual explanations for the behavior of RL agents. We first propose and implement a set of RL-specific counterfactual properties that ensure easily reachable counterfactuals with highly probable desired outcomes. We use a heuristic tree search of the agent's execution trajectories to find the most suitable counterfactuals based on the defined properties. We evaluate RACCER in two tasks as well as conduct a user study to show that RL-specific counterfactuals help users better understand agents' behavior compared to the current state-of-the-art approaches.

Jasmina Gajcin, Ivana Dusparic

Despite notable results in various fields over the recent years, deep reinforcement learning (DRL) algorithms lack transparency, affecting user trust and hindering their deployment to high-risk tasks. Causal confusion refers to a phenomenon where an agent learns spurious correlations between features which might not hold across the entire state space, preventing safe deployment to real tasks where such correlations might be broken. In this work, we examine whether an agent relies on spurious correlations in critical states, and propose an alternative subset of features on which it should base its decisions instead, to make it less susceptible to causal confusion. Our goal is to increase transparency of DRL agents by exposing the influence of learned spurious correlations on its decisions, and offering advice to developers about feature selection in different parts of state space, to avoid causal confusion. We propose ReCCoVER, an algorithm which detects causal confusion in agent's reasoning before deployment, by executing its policy in alternative environments where certain correlations between features do not hold. We demonstrate our approach in taxi and grid world environments, where ReCCoVER detects states in which an agent relies on spurious correlations and offers a set of features that should be considered instead.

Tom He, Jasmina Gajcin, Ivana Dusparic

Reinforcement learning (RL) is used in various robotic applications. RL enables agents to learn tasks autonomously by interacting with the environment. The more critical the tasks are, the higher the demand for the robustness of the RL systems. Causal RL combines RL and causal inference to make RL more robust. Causal RL agents use a causal representation to capture the invariant causal mechanisms that can be transferred from one task to another. Currently, there is limited research in Causal RL, and existing solutions are usually not complete or feasible for real-world applications. In this work, we propose CausalCF, the first complete Causal RL solution incorporating ideas from Causal Curiosity and CoPhy. Causal Curiosity provides an approach for using interventions, and CoPhy is modified to enable the RL agent to perform counterfactuals. Causal Curiosity has been applied to robotic grasping and manipulation tasks in CausalWorld. CausalWorld provides a realistic simulation environment based on the TriFinger robot. We apply CausalCF to complex robotic tasks and show that it improves the RL agent's robustness using CausalWorld.

Juan C. Rosero, Ivana Dusparic, Nicolás Cardozo

Reinforcement Learning (RL) is used extensively in Autonomous Systems (AS) as it enables learning at runtime without the need for a model of the environment or predefined actions. However, most applications of RL in AS, such as those based on Q-learning, can only optimize one objective, making it necessary in multi-objective systems to combine multiple objectives in a single objective function with predefined weights. A number of Multi-Objective Reinforcement Learning (MORL) techniques exist but they have mostly been applied in RL benchmarks rather than real-world AS systems. In this work, we use a MORL technique called Deep W-Learning (DWN) and apply it to the Emergent Web Servers exemplar, a self-adaptive server, to find the optimal configuration for runtime performance optimization. We compare DWN to two single-objective optimization implementations: ε-greedy algorithm and Deep Q-Networks. Our initial evaluation shows that DWN optimizes multiple objectives simultaneously with similar results than DQN and ε-greedy approaches, having a better performance for some metrics, and avoids issues associated with combining multiple objectives into a single utility function.

James McCarthy, Rahul Nair, Elizabeth Daly et al.

Explainability of Reinforcement Learning (RL) policies remains a challenging research problem, particularly when considering RL in a safety context. Understanding the decisions and intentions of an RL policy offer avenues to incorporate safety into the policy by limiting undesirable actions. We propose the use of a Boolean Decision Rules model to create a post-hoc rule-based summary of an agent's policy. We evaluate our proposed approach using a DQN agent trained on an implementation of a lava gridworld and show that, given a hand-crafted feature representation of this gridworld, simple generalised rules can be created, giving a post-hoc explainable summary of the agent's policy. We discuss possible avenues to introduce safety into a RL agent's policy by using rules generated by this rule-based model as constraints imposed on the agent's policy, as well as discuss how creating simple rule summaries of an agent's policy may help in the debugging process of RL agents.

Babatunji Omoniwa, Boris Galkin, Ivana Dusparic

Unmanned aerial vehicles (UAVs) serving as aerial base stations can be deployed to provide wireless connectivity to mobile users, such as vehicles. However, the density of vehicles on roads often varies spatially and temporally primarily due to mobility and traffic situations in a geographical area, making it difficult to provide ubiquitous service. Moreover, as energy-constrained UAVs hover in the sky while serving mobile users, they may be faced with interference from nearby UAV cells or other access points sharing the same frequency band, thereby impacting the system's energy efficiency (EE). Recent multi-agent reinforcement learning (MARL) approaches applied to optimise the users' coverage worked well in reasonably even densities but might not perform as well in uneven users' distribution, i.e., in urban road networks with uneven concentration of vehicles. In this work, we propose a density-aware communication-enabled multi-agent decentralised double deep Q-network (DACEMAD-DDQN) approach that maximises the total system's EE by jointly optimising the trajectory of each UAV, the number of connected users, and the UAVs' energy consumption while keeping track of dense and uneven users' distribution. Our result outperforms state-of-the-art MARL approaches in terms of EE by as much as 65% - 85%.

Alberto Castagna, Ivana Dusparic

Transfer learning in Reinforcement Learning (RL) has been widely studied to overcome training issues of Deep-RL, i.e., exploration cost, data availability and convergence time, by introducing a way to enhance training phase with external knowledge. Generally, knowledge is transferred from expert-agents to novices. While this fixes the issue for a novice agent, a good understanding of the task on expert agent is required for such transfer to be effective. As an alternative, in this paper we propose Expert-Free Online Transfer Learning (EF-OnTL), an algorithm that enables expert-free real-time dynamic transfer learning in multi-agent system. No dedicated expert exists, and transfer source agent and knowledge to be transferred are dynamically selected at each transfer step based on agents' performance and uncertainty. To improve uncertainty estimation, we also propose State Action Reward Next-State Random Network Distillation (sars-RND), an extension of RND that estimates uncertainty from RL agent-environment interaction. We demonstrate EF-OnTL effectiveness against a no-transfer scenario and advice-based baselines, with and without expert agents, in three benchmark tasks: Cart-Pole, a grid-based Multi-Team Predator-Prey (mt-pp) and Half Field Offense (HFO). Our results show that EF-OnTL achieve overall comparable performance when compared against advice-based baselines while not requiring any external input nor threshold tuning. EF-OnTL outperforms no-transfer with an improvement related to the complexity of the task addressed.

Jernej Hribar, Luke Hackett, Ivana Dusparic

In this paper, we build on advances introduced by the Deep Q-Networks (DQN) approach to extend the multi-objective tabular Reinforcement Learning (RL) algorithm W-learning to large state spaces. W-learning algorithm can naturally solve the competition between multiple single policies in multi-objective environments. However, the tabular version does not scale well to environments with large state spaces. To address this issue, we replace underlying Q-tables with DQN, and propose an addition of W-Networks, as a replacement for tabular weights (W) representations. We evaluate the resulting Deep W-Networks (DWN) approach in two widely-accepted multi-objective RL benchmarks: deep sea treasure and multi-objective mountain car. We show that DWN solves the competition between multiple policies while outperforming the baseline in the form of a DQN solution. Additionally, we demonstrate that the proposed algorithm can find the Pareto front in both tested environments.

Jasmina Gajcin, Jovan Jeromela, Ivana Dusparic

Reinforcement Learning (RL) is a learning paradigm in which the agent learns from its environment through trial and error. Deep reinforcement learning (DRL) algorithms represent the agent's policies using neural networks, making their decisions difficult to interpret. Explaining the behaviour of DRL agents is necessary to advance user trust, increase engagement, and facilitate integration with real-life tasks. Semifactual explanations aim to explain an outcome by providing "even if" scenarios, such as "even if the car were moving twice as slowly, it would still have to swerve to avoid crashing". Semifactuals help users understand the effects of different factors on the outcome and support the optimisation of resources. While extensively studied in psychology and even utilised in supervised learning, semifactuals have not been used to explain the decisions of RL systems. In this work, we develop a first approach to generating semifactual explanations for RL agents. We start by defining five properties of desirable semifactual explanations in RL and then introducing SGRL-Rewind and SGRL-Advance, the first algorithms for generating semifactual explanations in RL. We evaluate the algorithms in two standard RL environments and find that they generate semifactuals that are easier to reach, represent the agent's policy better, and are more diverse compared to baselines. Lastly, we conduct and analyse a user study to assess the participant's perception of semifactual explanations of the agent's actions.

Krešimir Kušić, Vinny Cahill, Ivana Dusparic

The operational effectiveness of digital-twin technology in motorway traffic management depends on the availability of a continuous flow of high-resolution real-time traffic data. To function as a proactive decision-making support layer within traffic management, a digital twin must also incorporate predicted traffic conditions in addition to real-time observations. Due to the spatio-temporal complexity and the time-variant, non-linear nature of traffic dynamics, predicting motorway traffic remains a difficult problem. Sequence-based deep-learning models offer clear advantages over classical machine learning and statistical models in capturing long-range, temporal dependencies in time-series traffic data, yet limitations in forecasting accuracy and model complexity point to the need for further improvements. To improve motorway traffic forecasting, this paper introduces a Geographically-aware Transformer-based Traffic Forecasting GATTF model, which exploits the geographical relationships between distributed sensors using their mutual information (MI). The model has been evaluated using real-time data from the Geneva motorway network in Switzerland and results confirm that incorporating geographical awareness through MI enhances the accuracy of GATTF forecasting compared to a standard Transformer, without increasing model complexity.

Raul de la Rosa, Ivana Dusparic, Nicolas Cardozo

Reinforcement Learning (RL) agents often struggle in real-world applications where environmental conditions are non-stationary, particularly when reward functions shift or the available action space expands. This paper introduces MORPHIN, a self-adaptive Q-learning framework that enables on-the-fly adaptation without full retraining. By integrating concept drift detection with dynamic adjustments to learning and exploration hyperparameters, MORPHIN adapts agents to changes in both the reward function and on-the-fly expansions of the agent's action space, while preserving prior policy knowledge to prevent catastrophic forgetting. We validate our approach using a Gridworld benchmark and a traffic signal control simulation. The results demonstrate that MORPHIN achieves superior convergence speed and continuous adaptation compared to a standard Q-learning baseline, improving learning efficiency by up to 1.7x.

Wenlong Wang, Ivana Dusparic, Yucheng Shi et al.

Model-based reinforcement learning (RL) offers a solution to the data inefficiency that plagues most model-free RL algorithms. However, learning a robust world model often requires complex and deep architectures, which are computationally expensive and challenging to train. Within the world model, sequence models play a critical role in accurate predictions, and various architectures have been explored, each with its own challenges. Currently, recurrent neural network (RNN)-based world models struggle with vanishing gradients and capturing long-term dependencies. Transformers, on the other hand, suffer from the quadratic memory and computational complexity of self-attention mechanisms, scaling as $O(n^2)$, where $n$ is the sequence length. To address these challenges, we propose a state space model (SSM)-based world model, Drama, specifically leveraging Mamba, that achieves $O(n)$ memory and computational complexity while effectively capturing long-term dependencies and enabling efficient training with longer sequences. We also introduce a novel sampling method to mitigate the suboptimality caused by an incorrect world model in the early training stages. Combining these techniques, Drama achieves a normalised score on the Atari100k benchmark that is competitive with other state-of-the-art (SOTA) model-based RL algorithms, using only a 7 million-parameter world model. Drama is accessible and trainable on off-the-shelf hardware, such as a standard laptop. Our code is available at https://github.com/realwenlongwang/Drama.git.

Shivani Tomar, Seshu Tirupathi, Elizabeth Daly et al.

Time series foundation models have recently gained a lot of attention due to their ability to model complex time series data encompassing different domains including traffic, energy, and weather. Although they exhibit strong average zero-shot performance on forecasting tasks, their predictions on certain critical regions of the data are not always reliable, limiting their usability in real-world applications, especially when data exhibits unique trends. In this paper, we propose a selective forecasting framework to identify these critical segments of time series using shapelets. We learn shapelets using shift-invariant dictionary learning on the validation split of the target domain dataset. Utilizing distance-based similarity to these shapelets, we facilitate the user to selectively discard unreliable predictions and be informed of the model's realistic capabilities. Empirical results on diverse benchmark time series datasets demonstrate that our approach leveraging both zero-shot and full-shot fine-tuned models reduces the overall error by an average of 22.17% for zero-shot and 22.62% for full-shot fine-tuned model. Furthermore, our approach using zero-shot and full-shot fine-tuned models, also outperforms its random selection counterparts by up to 21.41% and 21.43% on one of the datasets.

Dilina Rajapakse, Juan C. Rosero, Ivana Dusparic

Reinforcement Learning (RL) has demonstrated its ability to solve complex decision-making problems in a variety of domains, by optimizing reward signals obtained through interaction with an environment. However, many real-world scenarios involve multiple, potentially conflicting objectives that cannot be easily represented by a single scalar reward. Multi-Objective Reinforcement Learning (MORL) addresses this limitation by enabling agents to optimize several objectives simultaneously, explicitly reasoning about trade-offs between them. However, the ``black box" nature of the RL models makes the decision process behind chosen objective trade-offs unclear. Current Explainable Reinforcement Learning (XRL) methods are typically designed for single scalar rewards and do not account for explanations with respect to distinct objectives or user preferences. To address this gap, in this paper we propose TREX, a Trajectory based Explainability framework to explain Multi-objective Reinforcement Learning policies, based on trajectory attribution. TREX generates trajectories directly from the learned expert policy, across different user preferences and clusters them into semantically meaningful temporal segments. We quantify the influence of these behavioural segments on the Pareto trade-off by training complementary policies that exclude specific clusters, measuring the resulting relative deviation on the observed rewards and actions compared to the original expert policy. Experiments on multi-objective MuJoCo environments - HalfCheetah, Ant and Swimmer, demonstrate the framework's ability to isolate and quantify the specific behavioural patterns.

Chenyang Zhao, Vinny Cahill, Ivana Dusparic

Reward design has been one of the central challenges for real world reinforcement learning (RL) deployment, especially in settings with multiple objectives. Preference-based RL offers an appealing alternative by learning from human preferences over pairs of behavioural outcomes. More recently, RL from AI feedback (RLAIF) has demonstrated that large language models (LLMs) can generate preference labels at scale, mitigating the reliance on human annotators. However, existing RLAIF work typically focuses only on single-objective tasks, leaving the open question of how RLAIF handles systems that involve multiple objectives. In such systems trade-offs among conflicting objectives are difficult to specify, and policies risk collapsing into optimizing for a dominant goal. In this paper, we explore the extension of the RLAIF paradigm to multi-objective self-adaptive systems. We show that multi-objective RLAIF can produce policies that yield balanced trade-offs reflecting different user priorities without laborious reward engineering. We argue that integrating RLAIF into multi-objective RL offers a scalable path toward user-aligned policy learning in domains with inherently conflicting objectives.

Jasmina Gajcin, Ivana Dusparic

Understanding how failure occurs and how it can be prevented in reinforcement learning (RL) is necessary to enable debugging, maintain user trust, and develop personalized policies. Counterfactual reasoning has often been used to assign blame and understand failure by searching for the closest possible world in which the failure is avoided. However, current counterfactual state explanations in RL can only explain an outcome using just the current state features and offer no actionable recourse on how a negative outcome could have been prevented. In this work, we propose ACTER (Actionable Counterfactual Sequences for Explaining Reinforcement Learning Outcomes), an algorithm for generating counterfactual sequences that provides actionable advice on how failure can be avoided. ACTER investigates actions leading to a failure and uses the evolutionary algorithm NSGA-II to generate counterfactual sequences of actions that prevent it with minimal changes and high certainty even in stochastic environments. Additionally, ACTER generates a set of multiple diverse counterfactual sequences that enable users to correct failure in the way that best fits their preferences. We also introduce three diversity metrics that can be used for evaluating the diversity of counterfactual sequences. We evaluate ACTER in two RL environments, with both discrete and continuous actions, and show that it can generate actionable and diverse counterfactual sequences. We conduct a user study to explore how explanations generated by ACTER help users identify and correct failure.

Kim N. Nolle, Ivana Dusparic, Rhodri Cusack et al.

Continual learning (CL) is a branch of machine learning that aims to enable agents to adapt and generalise previously learned abilities so that these can be reapplied to new tasks or environments. This is particularly useful in multi-task settings or in non-stationary environments, where the dynamics can change over time. This is particularly relevant in cyber-physical systems such as autonomous driving. However, despite recent advances in CL, successfully applying it to reinforcement learning (RL) is still an open problem. This paper highlights open challenges in continual RL (CRL) based on experiments in an autonomous driving environment. In this environment, the agent must learn to successfully park in four different scenarios corresponding to parking spaces oriented at varying angles. The agent is successively trained in these four scenarios one after another, representing a CL environment, using Proximal Policy Optimisation (PPO). These experiments exposed a number of open challenges in CRL: finding suitable abstractions of the environment, oversensitivity to hyperparameters, catastrophic forgetting, and efficient use of neural network capacity. Based on these identified challenges, we present open research questions that are important to be addressed for creating robust CRL systems. In addition, the identified challenges call into question the suitability of neural networks for CL. We also identify the need for interdisciplinary research, in particular between computer science and neuroscience.

Emran Yasser Moustafa, Ivana Dusparic

Autonomous vehicles have shown promising potential to be a groundbreaking technology for improving the safety of road users. For these vehicles, as well as many other safety-critical robotic technologies, to be deployed in real-world applications, we require algorithms that can generalize well to unseen scenarios and data. Model-based reinforcement learning algorithms (MBRL) have demonstrated state-of-the-art performance and data efficiency across a diverse set of domains. However, these algorithms have also shown susceptibility to changes in the environment and its transition dynamics. In this work, we explore the performance and generalization capabilities of MBRL algorithms for autonomous driving, specifically in the simulated autonomous racing environment, Roboracer (formerly F1Tenth). We frame the head-to-head racing task as a learning problem using contextual Markov decision processes and parameterize the driving behavior of the adversaries using the context of the episode, thereby also parameterizing the transition and reward dynamics. We benchmark the behavior of MBRL algorithms in this environment and propose a novel context-aware extension of the existing literature, cMask. We demonstrate that context-aware MBRL algorithms generalize better to out-of-distribution adversary behaviors relative to context-free approaches. We also demonstrate that cMask displays strong generalization capabilities, as well as further performance improvement relative to other context-aware MBRL approaches when racing against adversaries with in-distribution behaviors.

James McCarthy, Radu Marinescu, Elizabeth Daly et al.

Risk-averse Constrained Reinforcement Learning (RaCRL) aims to learn policies that minimise the likelihood of rare and catastrophic constraint violations caused by an environment's inherent randomness. In general, risk-aversion leads to conservative exploration of the environment which typically results in converging to sub-optimal policies that fail to adequately maximise reward or, in some cases, fail to achieve the goal. In this paper, we propose an exploration-based approach for RaCRL called Optimistic Risk-averse Actor Critic (ORAC), which constructs an exploratory policy by maximising a local upper confidence bound of the state-action reward value function whilst minimising a local lower confidence bound of the risk-averse state-action cost value function. Specifically, at each step, the weighting assigned to the cost value is increased or decreased if it exceeds or falls below the safety constraint value. This way the policy is encouraged to explore uncertain regions of the environment to discover high reward states whilst still satisfying the safety constraints. Our experimental results demonstrate that the ORAC approach prevents convergence to sub-optimal policies and improves significantly the reward-cost trade-off in various continuous control tasks such as Safety-Gymnasium and a complex building energy management environment CityLearn.

Yucheng Shi, Wenlong Wang, Xiaowen Tao et al.

Dynamic scheduling of access to shared resources by autonomous systems is a challenging problem, characterized as being NP-hard. The complexity of this task leads to a combinatorial explosion of possibilities in highly dynamic systems where arriving requests must be continuously scheduled subject to strong safety and time constraints. An example of such a system is an unsignalized intersection, where automated vehicles' access to potential conflict zones must be dynamically scheduled. In this paper, we apply Neural Monte Carlo Tree Search (NMCTS) to the challenging task of scheduling platoons of vehicles crossing unsignalized intersections. Crucially, we introduce a transformation model that maps successive sequences of potentially conflicting road-space reservation requests from platoons of vehicles into a series of board-game-like problems and use NMCTS to search for solutions representing optimal road-space allocation schedules in the context of past allocations. To optimize search, we incorporate a prioritized re-sampling method with parallel NMCTS (PNMCTS) to improve the quality of training data. To optimize training, a curriculum learning strategy is used to train the agent to schedule progressively more complex boards culminating in overlapping boards that represent busy intersections. In a busy single four-way unsignalized intersection simulation, PNMCTS solved 95\% of unseen scenarios, reducing crossing time by 43\% in light and 52\% in heavy traffic versus first-in, first-out control. In a 3x3 multi-intersection network, the proposed method maintained free-flow in light traffic when all intersections are under control of PNMCTS and outperformed state-of-the-art RL-based traffic-light controllers in average travel time by 74.5\% and total throughput by 16\% in heavy traffic.

Jasmina Gajcin, Rahul Nair, Tejaswini Pedapati et al.

In complex tasks where the reward function is not straightforward and consists of a set of objectives, multiple reinforcement learning (RL) policies that perform task adequately, but employ different strategies can be trained by adjusting the impact of individual objectives on reward function. Understanding the differences in strategies between policies is necessary to enable users to choose between offered policies, and can help developers understand different behaviors that emerge from various reward functions and training hyperparameters in RL systems. In this work we compare behavior of two policies trained on the same task, but with different preferences in objectives. We propose a method for distinguishing between differences in behavior that stem from different abilities from those that are a consequence of opposing preferences of two RL agents. Furthermore, we use only data on preference-based differences in order to generate contrasting explanations about agents' preferences. Finally, we test and evaluate our approach on an autonomous driving task and compare the behavior of a safety-oriented policy and one that prefers speed.

Alberto Castagna, Ivana Dusparic

Reinforcement learning (RL) has been used in a range of simulated real-world tasks, e.g., sensor coordination, traffic light control, and on-demand mobility services. However, real world deployments are rare, as RL struggles with dynamic nature of real world environments, requiring time for learning a task and adapting to changes in the environment. Transfer Learning (TL) can help lower these adaptation times. In particular, there is a significant potential of applying TL in multi-agent RL systems, where multiple agents can share knowledge with each other, as well as with new agents that join the system. To obtain the most from inter-agent transfer, transfer roles (i.e., determining which agents act as sources and which as targets), as well as relevant transfer content parameters (e.g., transfer size) should be selected dynamically in each particular situation. As a first step towards fully dynamic transfers, in this paper we investigate the impact of TL transfer parameters with fixed source and target roles. Specifically, we label every agent-environment interaction with agent's epistemic confidence, and we filter the shared examples using varying threshold levels and sample sizes. We investigate impact of these parameters in two scenarios, a standard predator-prey RL benchmark and a simulation of a ride-sharing system with 200 vehicle agents and 10,000 ride-requests.

Boris Galkin, Babatunji Omoniwa, Ivana Dusparic

Unmanned Aerial Vehicles (UAVs) promise to become an intrinsic part of next generation communications, as they can be deployed to provide wireless connectivity to ground users to supplement existing terrestrial networks. The majority of the existing research into the use of UAV access points for cellular coverage considers rotary-wing UAV designs (i.e. quadcopters). However, we expect fixed-wing UAVs to be more appropriate for connectivity purposes in scenarios where long flight times are necessary (such as for rural coverage), as fixed-wing UAVs rely on a more energy-efficient form of flight when compared to the rotary-wing design. As fixed-wing UAVs are typically incapable of hovering in place, their deployment optimisation involves optimising their individual flight trajectories in a way that allows them to deliver high quality service to the ground users in an energy-efficient manner. In this paper, we propose a multi-agent deep reinforcement learning approach to optimise the energy efficiency of fixed-wing UAV cellular access points while still allowing them to deliver high-quality service to users on the ground. In our decentralized approach, each UAV is equipped with a Dueling Deep Q-Network (DDQN) agent which can adjust the 3D trajectory of the UAV over a series of timesteps. By coordinating with their neighbours, the UAVs adjust their individual flight trajectories in a manner that optimises the total system energy efficiency. We benchmark the performance of our approach against a series of heuristic trajectory planning strategies, and demonstrate that our method can improve the system energy efficiency by as much as 70%.

Babatunji Omoniwa, Boris Galkin, Ivana Dusparic

Unmanned aerial vehicles serving as aerial base stations (UAV-BSs) can be deployed to provide wireless connectivity to ground devices in events of increased network demand, points-of-failure in existing infrastructure, or disasters. However, it is challenging to conserve the energy of UAVs during prolonged coverage tasks, considering their limited on-board battery capacity. Reinforcement learning-based (RL) approaches have been previously used to improve energy utilization of multiple UAVs, however, a central cloud controller is assumed to have complete knowledge of the end-devices' locations, i.e., the controller periodically scans and sends updates for UAV decision-making. This assumption is impractical in dynamic network environments with UAVs serving mobile ground devices. To address this problem, we propose a decentralized Q-learning approach, where each UAV-BS is equipped with an autonomous agent that maximizes the connectivity of mobile ground devices while improving its energy utilization. Experimental results show that the proposed design significantly outperforms the centralized approaches in jointly maximizing the number of connected ground devices and the energy utilization of the UAV-BSs.

Babatunji Omoniwa, Maxime Gueriau, Ivana Dusparic

Recent research has shown the potential of using available mobile fog devices (such as smartphones, drones, domestic and industrial robots) as relays to minimize communication outages between sensors and destination devices, where localized Internet-of-Things services (e.g., manufacturing process control, health and security monitoring) are delivered. However, these mobile relays deplete energy when they move and transmit to distant destinations. As such, power-control mechanisms and intelligent mobility of the relay devices are critical in improving communication performance and energy utilization. In this paper, we propose a Q-learning-based decentralized approach where each mobile fog relay agent (MFRA) is controlled by an autonomous agent which uses reinforcement learning to simultaneously improve communication performance and energy utilization. Each autonomous agent learns based on the feedback from the destination and its own energy levels whether to remain active and forward the message, or become passive for that transmission phase. We evaluate the approach by comparing with the centralized approach, and observe that with lesser number of MFRAs, our approach is able to ensure reliable delivery of data and reduce overall energy cost by 56.76\% -- 88.03\%.

Ivana Dusparic, Nicolas Cardozo

Self-adaptive systems continuously adapt to changes in their execution environment. Capturing all possible changes to define suitable behaviour beforehand is unfeasible, or even impossible in the case of unknown changes, hence human intervention may be required. We argue that adapting to unknown situations is the ultimate challenge for self-adaptive systems. Learning-based approaches are used to learn the suitable behaviour to exhibit in the case of unknown situations, to minimize or fully remove human intervention. While such approaches can, to a certain extent, generalize existing adaptations to new situations, there is a number of breakthroughs that need to be achieved before systems can adapt to general unknown and unforeseen situations. We posit the research directions that need to be explored to achieve unanticipated adaptation from the perspective of learning-based self-adaptive systems. At minimum, systems need to define internal representations of previously unseen situations on-the-fly, extrapolate the relationship to the previously encountered situations to evolve existing adaptations, and reason about the feasibility of achieving their intrinsic goals in the new set of conditions. We close discussing whether, even when we can, we should indeed build systems that define their own behaviour and adapt their goals, without involving a human supervisor.

Nicolás Cardozo, Ivana Dusparic

Self-adaptive software systems continuously adapt in response to internal and external changes in their execution environment, captured as contexts. The COP paradigm posits a technique for the development of self-adaptive systems, capturing their main characteristics with specialized programming language constructs. COP adaptations are specified as independent modules composed in and out of the base system as contexts are activated and deactivated in response to sensed circumstances from the surrounding environment. However, the definition of adaptations, their contexts and associated specialized behavior, need to be specified at design time. In complex CPS this is intractable due to new unpredicted operating conditions. We propose Auto-COP, a new technique to enable generation of adaptations at run time. Auto-COP uses RL options to build action sequences, based on the previous instances of the system execution. Options are explored in interaction with the environment, and the most suitable options for each context are used to generate adaptations exploiting COP. To validate Auto-COP, we present two case studies exhibiting different system characteristics and application domains: a driving assistant and a robot delivery system. We present examples of Auto-COP code generated at run time, to illustrate the types of circumstances (contexts) requiring adaptation, and the corresponding generated adaptations for each context. We confirm that the generated adaptations exhibit correct system behavior measured by domain-specific performance metrics, while reducing the number of required execution/actuation steps by a factor of two showing that the adaptations are regularly selected by the running system as adaptive behavior is more appropriate than the execution of primitive actions.

Erika Fonseca, Boris Galkin, Ramy Amer et al.

Providing reliable connectivity to cellular-connected UAV can be very challenging; their performance highly depends on the nature of the surrounding environment, such as density and heights of the ground BSs. On the other hand, tall buildings might block undesired interference signals from ground BSs, thereby improving the connectivity between the UAVs and their serving BSs. To address the connectivity of UAVs in such environments, this paper proposes a RL algorithm to dynamically optimise the height of a UAV as it moves through the environment, with the goal of increasing the throughput or spectrum efficiency that it experiences. The proposed solution is evaluated in two settings: using a series of generated environments where we vary the number of BS and building densities, and in a scenario using real-world data obtained from an experiment in Dublin, Ireland. Results show that our proposed RL-based solution improves UAVs QoS by 6% to 41%, depending on the scenario. We also conclude that, when flying at heights higher than the buildings, building density variation has no impact on UAV QoS. On the other hand, BS density can negatively impact UAV QoS, with higher numbers of BSs generating more interference and deteriorating UAV performance.

Amit Prasad, Ivana Dusparic

Advances in renewable energy generation and introduction of the government targets to improve energy efficiency gave rise to a concept of a Zero Energy Building (ZEB). A ZEB is a building whose net energy usage over a year is zero, i.e., its energy use is not larger than its overall renewables generation. A collection of ZEBs forms a Zero Energy Community (ZEC). This paper addresses the problem of energy sharing in such a community. This is different from previously addressed energy sharing between buildings as our focus is on the improvement of community energy status, while traditionally research focused on reducing losses due to transmission and storage, or achieving economic gains. We model this problem in a multi-agent environment and propose a Deep Reinforcement Learning (DRL) based solution. Each building is represented by an intelligent agent that learns over time the appropriate behaviour to share energy. We have evaluated the proposed solution in a multi-agent simulation built using osBrain. Results indicate that with time agents learn to collaborate and learn a policy comparable to the optimal policy, which in turn improves the ZEC's energy status. Buildings with no renewables preferred to request energy from their neighbours rather than from the supply grid.