'Reincarnation' in reinforcement learning has been proposed as a formalisation of reusing prior computation from past experiments when training an agent in an environment. In this paper, we present a brief foray into the paradigm of reincarnation in the multi-agent (MA) context. We consider the case where only some agents are reincarnated, whereas the others are trained from scratch -- selective reincarnation. In the fully-cooperative MA setting with heterogeneous agents, we demonstrate that selective reincarnation can lead to higher returns than training fully from scratch, and faster convergence than training with full reincarnation. However, the choice of which agents to reincarnate in a heterogeneous system is vitally important to the outcome of the training -- in fact, a poor choice can lead to considerably worse results than the alternatives. We argue that a rich field of work exists here, and we hope that our effort catalyses further energy in bringing the topic of reincarnation to the multi-agent realm.

Sequential decision-making agents struggle with long horizon tasks, since solving them requires multi-step reasoning. Most reinforcement learning (RL) algorithms address this challenge by improved credit assignment, introducing memory capability, altering the agent's intrinsic motivation (i.e. exploration) or its worldview (i.e. knowledge representation). Many of these components could be learned from offline data. In this work, we follow the hypothesis that exploration and representation learning can be improved by separately learning two different models from a single offline dataset. We show that learning a state representation using noise-contrastive estimation and a model of auxiliary reward separately from a single collection of human demonstrations can significantly improve the sample efficiency on the challenging NetHack benchmark. We also ablate various components of our experimental setting and highlight crucial insights.

The sharing-economy-based business model has recently seen success in the transportation and accommodation sectors with companies like Uber and Airbnb. There is growing interest in applying this model to energy systems, with modalities like peer-to-peer (P2P) Energy Trading, Electric Vehicles (EV)-based Vehicle-to-Grid (V2G), Vehicle-to-Home (V2H), Vehicle-to-Vehicle (V2V), and Battery Swapping Technology (BST). In this work, we exploit the increasing diffusion of EVs to realize a crowdsourcing platform called e-Uber that jointly enables ride-sharing and energy-sharing through V2G and BST. e-Uber exploits spatial crowdsourcing, reinforcement learning, and reverse auction theory. Specifically, the platform uses reinforcement learning to understand the drivers' preferences towards different ride-sharing and energy-sharing tasks. Based on these preferences, a personalized list is recommended to each driver through CMAB-based Algorithm for task Recommendation System (CARS). Drivers bid on their preferred tasks in their list in a reverse auction fashion. Then e-Uber solves the task assignment optimization problem that minimizes cost and guarantees V2G energy requirement. We prove that this problem is NP-hard and introduce a bipartite matching-inspired heuristic, Bipartite Matching-based Winner selection (BMW), that has polynomial time complexity. Results from experiments using real data from NYC taxi trips and energy consumption show that e-Uber performs close to the optimum and finds better solutions compared to a state-of-the-art approach

Existing statistical methods can estimate a policy, or a mapping from covariates to decisions, which can then instruct decision makers (e.g., whether to administer hypotension treatment based on covariates blood pressure and heart rate). There is great interest in using such data-driven policies in healthcare. However, it is often important to explain to the healthcare provider, and to the patient, how a new policy differs from the current standard of care. This end is facilitated if one can pinpoint the aspects of the policy (i.e., the parameters for blood pressure and heart rate) that change when moving from the standard of care to the new, suggested policy. To this end, we adapt ideas from Trust Region Policy Optimization (TRPO). In our work, however, unlike in TRPO, the difference between the suggested policy and standard of care is required to be sparse, aiding with interpretability. This yields ``relative sparsity," where, as a function of a tuning parameter, $\lambda$, we can approximately control the number of parameters in our suggested policy that differ from their counterparts in the standard of care (e.g., heart rate only). We propose a criterion for selecting $\lambda$, perform simulations, and illustrate our method with a real, observational healthcare dataset, deriving a policy that is easy to explain in the context of the current standard of care. Our work promotes the adoption of data-driven decision aids, which have great potential to improve health outcomes.

Research in model-based reinforcement learning has made significant progress in recent years. Compared to single-agent settings, the exponential dimension growth of the joint state-action space in multi-agent systems dramatically increases the complexity of the environment dynamics, which makes it infeasible to learn an accurate global model and thus necessitates the use of agent-wise local models. However, during multi-step model rollouts, the prediction of one local model can affect the predictions of other local models in the next step. As a result, local prediction errors can be propagated to other localities and eventually give rise to considerably large global errors. Furthermore, since the models are generally used to predict for multiple steps, simply minimizing one-step prediction errors regardless of their long-term effect on other models may further aggravate the propagation of local errors. To this end, we propose Models as AGents (MAG), a multi-agent model optimization framework that reversely treats the local models as multi-step decision making agents and the current policies as the dynamics during the model rollout process. In this way, the local models are able to consider the multi-step mutual affect between each other before making predictions. Theoretically, we show that the objective of MAG is approximately equivalent to maximizing a lower bound of the true environment return. Experiments on the challenging StarCraft II benchmark demonstrate the effectiveness of MAG.

In this paper, a novel switching pushing skill algorithm is proposed to improve the efficiency of planar non-prehensile manipulation, which draws inspiration from human pushing actions and comprises two sub-problems, i.e., discrete decision-making of pushing point and continuous feedback control of pushing action. In order to solve the sub-problems above, a combination of Model Predictive Control (MPC) and Deep Reinforcement Learning (DRL) method is employed. Firstly, the selection of pushing point is modeled as a Markov decision process,and an off-policy DRL method is used by reshaping the reward function to train the decision-making model for selecting pushing point from a pre-constructed set based on the current state. Secondly, a motion constraint region (MCR) is constructed for the specific pushing point based on the distance from the target, followed by utilizing the MPC controller to regulate the motion of the object within the MCR towards the target pose. The trigger condition for switching the pushing point occurs when the object reaches the boundary of the MCR under the pushing action. Subsequently, the pushing point and the controller are updated iteratively until the target pose is reached. We conducted pushing experiments on four distinct object shapes in both simulated and physical environments to evaluate our method. The results indicate that our method achieves a significantly higher training efficiency, with a training time that is only about 20% of the baseline method while maintaining around the same success rate. Moreover, our method outperforms the baseline method in terms of both training and execution efficiency of pushing operations, allowing for rapid learning of robot pushing skills.

Several research works have applied Reinforcement Learning (RL) algorithms to solve the Rate Adaptation (RA) problem in Wi-Fi networks. The dynamic nature of the radio link requires the algorithms to be responsive to changes in link quality. Delays in the execution of the algorithm may be detrimental to its performance, which in turn may decrease network performance. This aspect has been overlooked in the state of the art. In this paper, we present an analysis of common computational delays in RL-based RA algorithms, and propose a methodology that may be applied to reduce these computational delays and increase the efficiency of this type of algorithms. We apply the proposed methodology to an existing RL-based RA algorithm. The obtained experimental results indicate a reduction of one order of magnitude in the execution time of the algorithm, improving its responsiveness to link quality changes.

Humans learn quickly even in tasks that contain complex visual information. This is due in part to the efficient formation of compressed representations of visual information, allowing for better generalization and robustness. However, compressed representations alone are insufficient for explaining the high speed of human learning. Reinforcement learning (RL) models that seek to replicate this impressive efficiency may do so through the use of factored representations of tasks. These informationally simplistic representations of tasks are similarly motivated as the use of compressed representations of visual information. Recent studies have connected biological visual perception to disentangled and compressed representations. This raises the question of how humans learn to efficiently represent visual information in a manner useful for learning tasks. In this paper we present a model of human factored representation learning based on an altered form of a $\beta$-Variational Auto-encoder used in a visual learning task. Modelling results demonstrate a trade-off in the informational complexity of model latent dimension spaces, between the speed of learning and the accuracy of reconstructions.

Deep learning-based approaches for generating novel drug molecules with specific properties have gained a lot of interest in the last years. Recent studies have demonstrated promising performance for string-based generation of novel molecules utilizing reinforcement learning. In this paper, we develop a unified framework for using reinforcement learning for de novo drug design, wherein we systematically study various on- and off-policy reinforcement learning algorithms and replay buffers to learn an RNN-based policy to generate novel molecules predicted to be active against the dopamine receptor DRD2. Our findings suggest that it is advantageous to use at least both top-scoring and low-scoring molecules for updating the policy when structural diversity is essential. Using all generated molecules at an iteration seems to enhance performance stability for on-policy algorithms. In addition, when replaying high, intermediate, and low-scoring molecules, off-policy algorithms display the potential of improving the structural diversity and number of active molecules generated, but possibly at the cost of a longer exploration phase. Our work provides an open-source framework enabling researchers to investigate various reinforcement learning methods for de novo drug design.

Offline reinforcement learning (RL) allows for the training of competent agents from offline datasets without any interaction with the environment. Online finetuning of such offline models can further improve performance. But how should we ideally finetune agents obtained from offline RL training? While offline RL algorithms can in principle be used for finetuning, in practice, their online performance improves slowly. In contrast, we show that it is possible to use standard online off-policy algorithms for faster improvement. However, we find this approach may suffer from policy collapse, where the policy undergoes severe performance deterioration during initial online learning. We investigate the issue of policy collapse and how it relates to data diversity, algorithm choices and online replay distribution. Based on these insights, we propose a conservative policy optimization procedure that can achieve stable and sample-efficient online learning from offline pretraining.