Thomy Phan

Fabian Ritz, Thomy Phan, Andreas Sedlmeier et al.

The development of Machine Learning (ML) models is more than just a special case of software development (SD): ML models acquire properties and fulfill requirements even without direct human interaction in a seemingly uncontrollable manner. Nonetheless, the underlying processes can be described in a formal way. We define a comprehensive SD process model for ML that encompasses most tasks and artifacts described in the literature in a consistent way. In addition to the production of the necessary artifacts, we also focus on generating and validating fitting descriptions in the form of specifications. We stress the importance of further evolving the ML model throughout its life-cycle even after initial training and testing. Thus, we provide various interaction points with standard SD processes in which ML often is an encapsulated task. Further, our SD process model allows to formulate ML as a (meta-) optimization problem. If automated rigorously, it can be used to realize self-adaptive autonomous systems. Finally, our SD process model features a description of time that allows to reason about the progress within ML development processes. This might lead to further applications of formal methods within the field of ML.

Michael Kölle, Felix Topp, Thomy Phan et al.

Multi-Agent Reinforcement Learning is becoming increasingly more important in times of autonomous driving and other smart industrial applications. Simultaneously a promising new approach to Reinforcement Learning arises using the inherent properties of quantum mechanics, reducing the trainable parameters of a model significantly. However, gradient-based Multi-Agent Quantum Reinforcement Learning methods often have to struggle with barren plateaus, holding them back from matching the performance of classical approaches. While gradient free Quantum Reinforcement Learning methods may alleviate some of these challenges, they too are not immune to the difficulties posed by barren plateaus. We build upon an existing approach for gradient free Quantum Reinforcement Learning and propose three genetic variations with Variational Quantum Circuits for Multi-Agent Reinforcement Learning using evolutionary optimization. We evaluate our genetic variations in the Coin Game environment and also compare them to classical approaches. We showed that our Variational Quantum Circuit approaches perform significantly better compared to a neural network with a similar amount of trainable parameters. Compared to the larger neural network, our approaches archive similar results using $97.88\%$ less parameters.

Philipp Altmann, Fabian Ritz, Leonard Feuchtinger et al.

The safe application of reinforcement learning (RL) requires generalization from limited training data to unseen scenarios. Yet, fulfilling tasks under changing circumstances is a key challenge in RL. Current state-of-the-art approaches for generalization apply data augmentation techniques to increase the diversity of training data. Even though this prevents overfitting to the training environment(s), it hinders policy optimization. Crafting a suitable observation, only containing crucial information, has been shown to be a challenging task itself. To improve data efficiency and generalization capabilities, we propose Compact Reshaped Observation Processing (CROP) to reduce the state information used for policy optimization. By providing only relevant information, overfitting to a specific training layout is precluded and generalization to unseen environments is improved. We formulate three CROPs that can be applied to fully observable observation- and action-spaces and provide methodical foundation. We empirically show the improvements of CROP in a distributionally shifted safety gridworld. We furthermore provide benchmark comparisons to full observability and data-augmentation in two different-sized procedurally generated mazes.

Michael Kölle, Karola Schneider, Sabrina Egger et al.

In recent years, Multi-Agent Reinforcement Learning (MARL) has found application in numerous areas of science and industry, such as autonomous driving, telecommunications, and global health. Nevertheless, MARL suffers from, for instance, an exponential growth of dimensions. Inherent properties of quantum mechanics help to overcome these limitations, e.g., by significantly reducing the number of trainable parameters. Previous studies have developed an approach that uses gradient-free quantum Reinforcement Learning and evolutionary optimization for variational quantum circuits (VQCs) to reduce the trainable parameters and avoid barren plateaus as well as vanishing gradients. This leads to a significantly better performance of VQCs compared to classical neural networks with a similar number of trainable parameters and a reduction in the number of parameters by more than 97 \% compared to similarly good neural networks. We extend an approach of Kölle et al. by proposing a Gate-Based, a Layer-Based, and a Prototype-Based concept to mutate and recombine VQCs. Our results show the best performance for mutation-only strategies and the Gate-Based approach. In particular, we observe a significantly better score, higher total and own collected coins, as well as a superior own coin rate for the best agent when evaluated in the Coin Game environment.

Philipp Altmann, Thomy Phan, Fabian Ritz et al.

We propose discriminative reward co-training (DIRECT) as an extension to deep reinforcement learning algorithms. Building upon the concept of self-imitation learning (SIL), we introduce an imitation buffer to store beneficial trajectories generated by the policy determined by their return. A discriminator network is trained concurrently to the policy to distinguish between trajectories generated by the current policy and beneficial trajectories generated by previous policies. The discriminator's verdict is used to construct a reward signal for optimizing the policy. By interpolating prior experience, DIRECT is able to act as a surrogate, steering policy optimization towards more valuable regions of the reward landscape thus learning an optimal policy. Our results show that DIRECT outperforms state-of-the-art algorithms in sparse- and shifting-reward environments being able to provide a surrogate reward to the policy and direct the optimization towards valuable areas.

Thomy Phan, Benran Zhang, Shao-Hung Chan et al.

Anytime multi-agent path finding (MAPF) is a promising approach to scalable path optimization in multi-agent systems. MAPF-LNS, based on Large Neighborhood Search (LNS), is the current state-of-the-art approach where a fast initial solution is iteratively optimized by destroying and repairing selected paths of the solution. Current MAPF-LNS variants commonly use an adaptive selection mechanism to choose among multiple destroy heuristics. However, to determine promising destroy heuristics, MAPF-LNS requires a considerable amount of exploration time. As common destroy heuristics are non-adaptive, any performance bottleneck caused by these heuristics cannot be overcome via adaptive heuristic selection alone, thus limiting the overall effectiveness of MAPF-LNS in terms of solution cost. In this paper, we propose Adaptive Delay-based Destroy-and-Repair Enhanced with Success-based Self-Learning (ADDRESS) as a single-destroy-heuristic variant of MAPF-LNS. ADDRESS applies restricted Thompson Sampling to the top-K set of the most delayed agents to select a seed agent for adaptive LNS neighborhood generation. We evaluate ADDRESS in multiple maps from the MAPF benchmark set and demonstrate cost improvements by at least 50% in large-scale scenarios with up to a thousand agents, compared with the original MAPF-LNS and other state-of-the-art methods.

Michael Kölle, Yannick Erpelding, Fabian Ritz et al.

Recent advances in Multi-Agent Reinforcement Learning have prompted the modeling of intricate interactions between agents in simulated environments. In particular, the predator-prey dynamics have captured substantial interest and various simulations been tailored to unique requirements. To prevent further time-intensive developments, we introduce Aquarium, a comprehensive Multi-Agent Reinforcement Learning environment for predator-prey interaction, enabling the study of emergent behavior. Aquarium is open source and offers a seamless integration of the PettingZoo framework, allowing a quick start with proven algorithm implementations. It features physics-based agent movement on a two-dimensional, edge-wrapping plane. The agent-environment interaction (observations, actions, rewards) and the environment settings (agent speed, prey reproduction, predator starvation, and others) are fully customizable. Besides a resource-efficient visualization, Aquarium supports to record video files, providing a visual comprehension of agent behavior. To demonstrate the environment's capabilities, we conduct preliminary studies which use PPO to train multiple prey agents to evade a predator. In accordance to the literature, we find Individual Learning to result in worse performance than Parameter Sharing, which significantly improves coordination and sample-efficiency.

Thomy Phan, Taoan Huang, Bistra Dilkina et al.

Anytime multi-agent path finding (MAPF) is a promising approach to scalable path optimization in large-scale multi-agent systems. State-of-the-art anytime MAPF is based on Large Neighborhood Search (LNS), where a fast initial solution is iteratively optimized by destroying and repairing a fixed number of parts, i.e., the neighborhood, of the solution, using randomized destroy heuristics and prioritized planning. Despite their recent success in various MAPF instances, current LNS-based approaches lack exploration and flexibility due to greedy optimization with a fixed neighborhood size which can lead to low quality solutions in general. So far, these limitations have been addressed with extensive prior effort in tuning or offline machine learning beyond actual planning. In this paper, we focus on online learning in LNS and propose Bandit-based Adaptive LArge Neighborhood search Combined with Exploration (BALANCE). BALANCE uses a bi-level multi-armed bandit scheme to adapt the selection of destroy heuristics and neighborhood sizes on the fly during search. We evaluate BALANCE on multiple maps from the MAPF benchmark set and empirically demonstrate cost improvements of at least 50% compared to state-of-the-art anytime MAPF in large-scale scenarios. We find that Thompson Sampling performs particularly well compared to alternative multi-armed bandit algorithms.

Philipp Altmann, Jonas Stein, Michael Kölle et al.

Quantum computing (QC) in the current NISQ era is still limited in size and precision. Hybrid applications mitigating those shortcomings are prevalent to gain early insight and advantages. Hybrid quantum machine learning (QML) comprises both the application of QC to improve machine learning (ML) and ML to improve QC architectures. This work considers the latter, leveraging reinforcement learning (RL) to improve quantum circuit design (QCD), which we formalize by a set of generic objectives. Furthermore, we propose qcd-gym, a concrete framework formalized as a Markov decision process, to enable learning policies capable of controlling a universal set of continuously parameterized quantum gates. Finally, we provide benchmark comparisons to assess the shortcomings and strengths of current state-of-the-art RL algorithms.

Robert Müller, Hasan Turalic, Thomy Phan et al.

In the realm of Multi-Agent Reinforcement Learning (MARL), prevailing approaches exhibit shortcomings in aligning with human learning, robustness, and scalability. Addressing this, we introduce ClusterComm, a fully decentralized MARL framework where agents communicate discretely without a central control unit. ClusterComm utilizes Mini-Batch-K-Means clustering on the last hidden layer's activations of an agent's policy network, translating them into discrete messages. This approach outperforms no communication and competes favorably with unbounded, continuous communication and hence poses a simple yet effective strategy for enhancing collaborative task-solving in MARL.

Shao-Hung Chan, Thomy Phan, Jiaoyang Li et al.

Multi-Agent Path Finding (MAPF) is the problem of finding a set of collision-free paths, one for each agent in a shared environment. Its objective is to minimize the sum of path costs (SOC), where the path cost of each agent is defined as the travel time from its start location to its target location. Explicit Estimation Conflict-Based Search (EECBS) is the leading algorithm for bounded-suboptimal MAPF, with the SOC of the solution being at most a user-specified factor $w$ away from optimal. EECBS maintains sets of paths and a lower bound $LB$ on the optimal SOC. Then, it iteratively selects a set of paths whose SOC is at most $w \cdot LB$ and introduces constraints to resolve collisions. For each path in a set, EECBS maintains a lower bound on its optimal path that satisfies constraints. By finding an individually bounded-suboptimal path with cost at most a threshold of $w$ times its lower bound, EECBS guarantees to find a bounded-suboptimal solution. To speed up EECBS, previous work uses flex distribution to increase the threshold. Though EECBS with flex distribution guarantees to find a bounded-suboptimal solution, increasing the thresholds may push the SOC beyond $w \cdot LB$, forcing EECBS to switch among different sets of paths instead of resolving collisions on a particular set of paths, and thus reducing efficiency. To address this issue, we propose Conflict-Based Flex Distribution that distributes flex in proportion to the number of collisions. We also estimate the delays needed to satisfy constraints and propose Delay-Based Flex Distribution. On top of that, we propose Mixed-Strategy Flex Distribution, combining both in a hierarchical framework. We prove that EECBS with our new flex distribution mechanisms is complete and bounded-suboptimal. Our experiments show that our approaches outperform the original (greedy) flex distribution.

Fabian Ritz, Thomy Phan, Robert Müller et al.

A characteristic of reinforcement learning is the ability to develop unforeseen strategies when solving problems. While such strategies sometimes yield superior performance, they may also result in undesired or even dangerous behavior. In industrial scenarios, a system's behavior also needs to be predictable and lie within defined ranges. To enable the agents to learn (how) to align with a given specification, this paper proposes to explicitly transfer functional and non-functional requirements into shaped rewards. Experiments are carried out on the smart factory, a multi-agent environment modeling an industrial lot-size-one production facility, with up to eight agents and different multi-agent reinforcement learning algorithms. Results indicate that compliance with functional and non-functional constraints can be achieved by the proposed approach.

Markus Friedrich, Sebastian Feld, Thomy Phan et al.

Extracting a Construction Tree from potentially noisy point clouds is an important aspect of Reverse Engineering tasks in Computer Aided Design. Solutions based on algorithmic geometry impose constraints on usable model representations (e.g. quadric surfaces only) and noise robustness. Re-formulating the problem as a combinatorial optimization problem and solving it with an Evolutionary Algorithm can mitigate some of these constraints at the cost of increased computational complexity. This paper proposes a graph-based search space partitioning scheme that is able to accelerate Evolutionary Construction Tree extraction while exploiting parallelization capabilities of modern CPUs. The evaluation indicates a speed-up up to a factor of $46.6$ compared to the baseline approach while resulting tree sizes increased by $25.2\%$ to $88.6\%$.

Thomas Gabor, Sebastian Feld, Hila Safi et al.

Current hardware limitations restrict the potential when solving quadratic unconstrained binary optimization (QUBO) problems via the quantum approximate optimization algorithm (QAOA) or quantum annealing (QA). Thus, we consider training neural networks in this context. We first discuss QUBO problems that originate from translated instances of the traveling salesman problem (TSP): Analyzing this representation via autoencoders shows that there is way more information included than necessary to solve the original TSP. Then we show that neural networks can be used to solve TSP instances from both QUBO input and autoencoders' hiddenstate representation. We finally generalize the approach and successfully train neural networks to solve arbitrary QUBO problems, sketching means to use neuromorphic hardware as a simulator or an additional co-processor for quantum computing.

Thomas Gabor, Leo Sünkel, Fabian Ritz et al.

We discuss the synergetic connection between quantum computing and artificial intelligence. After surveying current approaches to quantum artificial intelligence and relating them to a formal model for machine learning processes, we deduce four major challenges for the future of quantum artificial intelligence: (i) Replace iterative training with faster quantum algorithms, (ii) distill the experience of larger amounts of data into the training process, (iii) allow quantum and classical components to be easily combined and exchanged, and (iv) build tools to thoroughly analyze whether observed benefits really stem from quantum properties of the algorithm.

Andreas Sedlmeier, Thomas Gabor, Thomy Phan et al.

Robustness to out-of-distribution (OOD) data is an important goal in building reliable machine learning systems. Especially in autonomous systems, wrong predictions for OOD inputs can cause safety critical situations. As a first step towards a solution, we consider the problem of detecting such data in a value-based deep reinforcement learning (RL) setting. Modelling this problem as a one-class classification problem, we propose a framework for uncertainty-based OOD classification: UBOOD. It is based on the effect that an agent's epistemic uncertainty is reduced for situations encountered during training (in-distribution), and thus lower than for unencountered (OOD) situations. Being agnostic towards the approach used for estimating epistemic uncertainty, combinations with different uncertainty estimation methods, e.g. approximate Bayesian inference methods or ensembling techniques are possible. We further present a first viable solution for calculating a dynamic classification threshold, based on the uncertainty distribution of the training data. Evaluation shows that the framework produces reliable classification results when combined with ensemble-based estimators, while the combination with concrete dropout-based estimators fails to reliably detect OOD situations. In summary, UBOOD presents a viable approach for OOD classification in deep RL settings by leveraging the epistemic uncertainty of the agent's value function.

Thomy Phan, Thomas Gabor, Robert Müller et al.

We propose Stable Yet Memory Bounded Open-Loop (SYMBOL) planning, a general memory bounded approach to partially observable open-loop planning. SYMBOL maintains an adaptive stack of Thompson Sampling bandits, whose size is bounded by the planning horizon and can be automatically adapted according to the underlying domain without any prior domain knowledge beyond a generative model. We empirically test SYMBOL in four large POMDP benchmark problems to demonstrate its effectiveness and robustness w.r.t. the choice of hyperparameters and evaluate its adaptive memory consumption. We also compare its performance with other open-loop planning algorithms and POMCP.

Carsten Hahn, Thomy Phan, Thomas Gabor et al.

In nature, flocking or swarm behavior is observed in many species as it has beneficial properties like reducing the probability of being caught by a predator. In this paper, we propose SELFish (Swarm Emergent Learning Fish), an approach with multiple autonomous agents which can freely move in a continuous space with the objective to avoid being caught by a present predator. The predator has the property that it might get distracted by multiple possible preys in its vicinity. We show that this property in interaction with self-interested agents which are trained with reinforcement learning to solely survive as long as possible leads to flocking behavior similar to Boids, a common simulation for flocking behavior. Furthermore we present interesting insights in the swarming behavior and in the process of agents being caught in our modeled environment.

Thomy Phan, Lenz Belzner, Marie Kiermeier et al.

State-of-the-art approaches to partially observable planning like POMCP are based on stochastic tree search. While these approaches are computationally efficient, they may still construct search trees of considerable size, which could limit the performance due to restricted memory resources. In this paper, we propose Partially Observable Stacked Thompson Sampling (POSTS), a memory bounded approach to open-loop planning in large POMDPs, which optimizes a fixed size stack of Thompson Sampling bandits. We empirically evaluate POSTS in four large benchmark problems and compare its performance with different tree-based approaches. We show that POSTS achieves competitive performance compared to tree-based open-loop planning and offers a performance-memory tradeoff, making it suitable for partially observable planning with highly restricted computational and memory resources.

Thomy Phan, Kyrill Schmid, Lenz Belzner et al.

Decision making in multi-agent systems (MAS) is a great challenge due to enormous state and joint action spaces as well as uncertainty, making centralized control generally infeasible. Decentralized control offers better scalability and robustness but requires mechanisms to coordinate on joint tasks and to avoid conflicts. Common approaches to learn decentralized policies for cooperative MAS suffer from non-stationarity and lacking credit assignment, which can lead to unstable and uncoordinated behavior in complex environments. In this paper, we propose Strong Emergent Policy approximation (STEP), a scalable approach to learn strong decentralized policies for cooperative MAS with a distributed variant of policy iteration. For that, we use function approximation to learn from action recommendations of a decentralized multi-agent planning algorithm. STEP combines decentralized multi-agent planning with centralized learning, only requiring a generative model for distributed black box optimization. We experimentally evaluate STEP in two challenging and stochastic domains with large state and joint action spaces and show that STEP is able to learn stronger policies than standard multi-agent reinforcement learning algorithms, when combining multi-agent open-loop planning with centralized function approximation. The learned policies can be reintegrated into the multi-agent planning process to further improve performance.

Andreas Sedlmeier, Thomas Gabor, Thomy Phan et al.

We consider the problem of detecting out-of-distribution (OOD) samples in deep reinforcement learning. In a value based reinforcement learning setting, we propose to use uncertainty estimation techniques directly on the agent's value estimating neural network to detect OOD samples. The focus of our work lies in analyzing the suitability of approximate Bayesian inference methods and related ensembling techniques that generate uncertainty estimates. Although prior work has shown that dropout-based variational inference techniques and bootstrap-based approaches can be used to model epistemic uncertainty, the suitability for detecting OOD samples in deep reinforcement learning remains an open question. Our results show that uncertainty estimation can be used to differentiate in- from out-of-distribution samples. Over the complete training process of the reinforcement learning agents, bootstrap-based approaches tend to produce more reliable epistemic uncertainty estimates, when compared to dropout-based approaches.

Thomas Gabor, Lenz Belzner, Thomy Phan et al.

As automatic optimization techniques find their way into industrial applications, the behavior of many complex systems is determined by some form of planner picking the right actions to optimize a given objective function. In many cases, the mapping of plans to objective reward may change due to unforeseen events or circumstances in the real world. In those cases, the planner usually needs some additional effort to adjust to the changed situation and reach its previous level of performance. Whenever we still need to continue polling the planner even during re-planning, it oftentimes exhibits severely lacking performance. In order to improve the planner's resilience to unforeseen change, we argue that maintaining a certain level of diversity amongst the considered plans at all times should be added to the planner's objective. Effectively, we encourage the planner to keep alternative plans to its currently best solution. As an example case, we implement a diversity-aware genetic algorithm using two different metrics for diversity (differing in their generality) and show that the blow in performance due to unexpected change can be severely lessened in the average case. We also analyze the parameter settings necessary for these techniques in order to gain an intuition how they can be incorporated into larger frameworks or process models for software and systems engineering.