This application paper explores the potential of using reinforcement learning (RL) to address the demands of Industry 4.0, including shorter time-to-market, mass customization, and batch size one production. Specifically, we present a use case in which the task is to transport and assemble goods through a model factory following predefined rules. Each simulation run involves placing a specific number of goods of random color at the entry point. The objective is to transport the goods to the assembly station, where two rivets are installed in each product, connecting the upper part to the lower part. Following the installation of rivets, blue products must be transported to the exit, while green products are to be transported to storage. The study focuses on the application of reinforcement learning techniques to address this problem and improve the efficiency of the production process.

Industry 4.0 is driven by demands like shorter time-to-market, mass customization of products, and batch size one production. Reinforcement Learning (RL), a machine learning paradigm shown to possess a great potential in improving and surpassing human level performance in numerous complex tasks, allows coping with the mentioned demands. In this paper, we present an OPC UA based Operational Technology (OT)-aware RL architecture, which extends the standard RL setting, combining it with the setting of digital twins. Moreover, we define an OPC UA information model allowing for a generalized plug-and-play like approach for exchanging the RL agent used. In conclusion, we demonstrate and evaluate the architecture, by creating a proof of concept. By means of solving a toy example, we show that this architecture can be used to determine the optimal policy using a real control system.

Granular materials are of critical interest to many robotic tasks in planetary science, construction, and manufacturing. However, the dynamics of granular materials are complex and often computationally very expensive to simulate. We propose a set of methodologies and a system for the fast simulation of granular materials on Graphics Processing Units (GPUs), and show that this simulation is fast enough for basic training with Reinforcement Learning algorithms, which currently require many dynamics samples to achieve acceptable performance. Our method models granular material dynamics using implicit timestepping methods for multibody rigid contacts, as well as algorithmic techniques for efficient parallel collision detection between pairs of particles and between particle and arbitrarily shaped rigid bodies, and programming techniques for minimizing warp divergence on Single-Instruction, Multiple-Thread (SIMT) chip architectures. We showcase our simulation system on several environments targeted toward robotic tasks, and release our simulator as an open-source tool.

In order to improve reproducibility, deep reinforcement learning (RL) has been adopting better scientific practices such as standardized evaluation metrics and reporting. However, the process of hyperparameter optimization still varies widely across papers, which makes it challenging to compare RL algorithms fairly. In this paper, we show that hyperparameter choices in RL can significantly affect the agent's final performance and sample efficiency, and that the hyperparameter landscape can strongly depend on the tuning seed which may lead to overfitting. We therefore propose adopting established best practices from AutoML, such as the separation of tuning and testing seeds, as well as principled hyperparameter optimization (HPO) across a broad search space. We support this by comparing multiple state-of-the-art HPO tools on a range of RL algorithms and environments to their hand-tuned counterparts, demonstrating that HPO approaches often have higher performance and lower compute overhead. As a result of our findings, we recommend a set of best practices for the RL community, which should result in stronger empirical results with fewer computational costs, better reproducibility, and thus faster progress. In order to encourage the adoption of these practices, we provide plug-and-play implementations of the tuning algorithms used in this paper at https://github.com/facebookresearch/how-to-autorl.

Multi-robot path finding in dynamic environments is a highly challenging classic problem. In the movement process, robots need to avoid collisions with other moving robots while minimizing their travel distance. Previous methods for this problem either continuously replan paths using heuristic search methods to avoid conflicts or choose appropriate collision avoidance strategies based on learning approaches. The former may result in long travel distances due to frequent replanning, while the latter may have low learning efficiency due to low sample exploration and utilization, and causing high training costs for the model. To address these issues, we propose a path planning method, MAPPOHR, which combines heuristic search, empirical rules, and multi-agent reinforcement learning. The method consists of two layers: a real-time planner based on the multi-agent reinforcement learning algorithm, MAPPO, which embeds empirical rules in the action output layer and reward functions, and a heuristic search planner used to create a global guiding path. During movement, the heuristic search planner replans new paths based on the instructions of the real-time planner. We tested our method in 10 different conflict scenarios. The experiments show that the planning performance of MAPPOHR is better than that of existing learning and heuristic methods. Due to the utilization of empirical knowledge and heuristic search, the learning efficiency of MAPPOHR is higher than that of existing learning methods.

Offline goal-conditioned RL (GCRL) offers a way to train general-purpose agents from fully offline datasets. In addition to being conservative within the dataset, the generalization ability to achieve unseen goals is another fundamental challenge for offline GCRL. However, to the best of our knowledge, this problem has not been well studied yet. In this paper, we study out-of-distribution (OOD) generalization of offline GCRL both theoretically and empirically to identify factors that are important. In a number of experiments, we observe that weighted imitation learning enjoys better generalization than pessimism-based offline RL method. Based on this insight, we derive a theory for OOD generalization, which characterizes several important design choices. We then propose a new offline GCRL method, Generalizable Offline goAl-condiTioned RL (GOAT), by combining the findings from our theoretical and empirical studies. On a new benchmark containing 9 independent identically distributed (IID) tasks and 17 OOD tasks, GOAT outperforms current state-of-the-art methods by a large margin.

Software developers spend a significant portion of time fixing bugs in their projects. To streamline this process, bug localization approaches have been proposed to identify the source code files that are likely responsible for a particular bug. Prior work proposed several similarity-based machine-learning techniques for bug localization. Despite significant advances in these techniques, they do not directly optimize the evaluation measures. We argue that directly optimizing evaluation measures can positively contribute to the performance of bug localization approaches. Therefore, In this paper, we utilize Reinforcement Learning (RL) techniques to directly optimize the ranking metrics. We propose RLocator, a Reinforcement Learning-based bug localization approach. We formulate RLocator using a Markov Decision Process (MDP) to optimize the evaluation measures directly. We present the technique and experimentally evaluate it based on a benchmark dataset of 8,316 bug reports from six highly popular Apache projects. The results of our evaluation reveal that RLocator achieves a Mean Reciprocal Rank (MRR) of 0.62, a Mean Average Precision (MAP) of 0.59, and a Top 1 score of 0.46. We compare RLocator with two state-of-the-art bug localization tools, FLIM and BugLocator. Our evaluation reveals that RLocator outperforms both approaches by a substantial margin, with improvements of 38.3% in MAP, 36.73% in MRR, and 23.68% in the Top K metric. These findings highlight that directly optimizing evaluation measures considerably contributes to performance improvement of the bug localization problem.

Current advances in recommender systems have been remarkably successful in optimizing immediate engagement. However, long-term user engagement, a more desirable performance metric, remains difficult to improve. Meanwhile, recent reinforcement learning (RL) algorithms have shown their effectiveness in a variety of long-term goal optimization tasks. For this reason, RL is widely considered as a promising framework for optimizing long-term user engagement in recommendation. Though promising, the application of RL heavily relies on well-designed rewards, but designing rewards related to long-term user engagement is quite difficult. To mitigate the problem, we propose a novel paradigm, recommender systems with human preferences (or Preference-based Recommender systems), which allows RL recommender systems to learn from preferences about users historical behaviors rather than explicitly defined rewards. Such preferences are easily accessible through techniques such as crowdsourcing, as they do not require any expert knowledge. With PrefRec, we can fully exploit the advantages of RL in optimizing long-term goals, while avoiding complex reward engineering. PrefRec uses the preferences to automatically train a reward function in an end-to-end manner. The reward function is then used to generate learning signals to train the recommendation policy. Furthermore, we design an effective optimization method for PrefRec, which uses an additional value function, expectile regression and reward model pre-training to improve the performance. We conduct experiments on a variety of long-term user engagement optimization tasks. The results show that PrefRec significantly outperforms previous state-of-the-art methods in all the tasks.

We study the problem of learning online packing skills for irregular 3D shapes, which is arguably the most challenging setting of bin packing problems. The goal is to consecutively move a sequence of 3D objects with arbitrary shapes into a designated container with only partial observations of the object sequence. Meanwhile, we take physical realizability into account, involving physics dynamics and constraints of a placement. The packing policy should understand the 3D geometry of the object to be packed and make effective decisions to accommodate it in the container in a physically realizable way. We propose a Reinforcement Learning (RL) pipeline to learn the policy. The complex irregular geometry and imperfect object placement together lead to huge solution space. Direct training in such space is prohibitively data intensive. We instead propose a theoretically-provable method for candidate action generation to reduce the action space of RL and the learning burden. A parameterized policy is then learned to select the best placement from the candidates. Equipped with an efficient method of asynchronous RL acceleration and a data preparation process of simulation-ready training sequences, a mature packing policy can be trained in a physics-based environment within 48 hours. Through extensive evaluation on a variety of real-life shape datasets and comparisons with state-of-the-art baselines, we demonstrate that our method outperforms the best-performing baseline on all datasets by at least 12.8% in terms of packing utility.

Optimizing noisy functions online, when evaluating the objective requires experiments on a deployed system, is a crucial task arising in manufacturing, robotics and many others. Often, constraints on safe inputs are unknown ahead of time, and we only obtain noisy information, indicating how close we are to violating the constraints. Yet, safety must be guaranteed at all times, not only for the final output of the algorithm. We introduce a general approach for seeking a stationary point in high dimensional non-linear stochastic optimization problems in which maintaining safety during learning is crucial. Our approach called LB-SGD is based on applying stochastic gradient descent (SGD) with a carefully chosen adaptive step size to a logarithmic barrier approximation of the original problem. We provide a complete convergence analysis of non-convex, convex, and strongly-convex smooth constrained problems, with first-order and zeroth-order feedback. Our approach yields efficient updates and scales better with dimensionality compared to existing approaches. We empirically compare the sample complexity and the computational cost of our method with existing safe learning approaches. Beyond synthetic benchmarks, we demonstrate the effectiveness of our approach on minimizing constraint violation in policy search tasks in safe reinforcement learning (RL).