Mobile Manipulation (MM) systems are ideal candidates for taking up the role of a personal assistant in unstructured real-world environments. Among other challenges, MM requires effective coordination of the robot's embodiments for executing tasks that require both mobility and manipulation. Reinforcement Learning (RL) holds the promise of endowing robots with adaptive behaviors, but most methods require prohibitively large amounts of data for learning a useful control policy. In this work, we study the integration of robotic reachability priors in actor-critic RL methods for accelerating the learning of MM for reaching and fetching tasks. Namely, we consider the problem of optimal base placement and the subsequent decision of whether to activate the arm for reaching a 6D target. For this, we devise a novel Hybrid RL method that handles discrete and continuous actions jointly, resorting to the Gumbel-Softmax reparameterization. Next, we train a reachability prior using data from the operational robot workspace, inspired by classical methods. Subsequently, we derive Boosted Hybrid RL (BHyRL), a novel algorithm for learning Q-functions by modeling them as a sum of residual approximators. Every time a new task needs to be learned, we can transfer our learned residuals and learn the component of the Q-function that is task-specific, hence, maintaining the task structure from prior behaviors. Moreover, we find that regularizing the target policy with a prior policy yields more expressive behaviors. We evaluate our method in simulation in reaching and fetching tasks of increasing difficulty, and we show the superior performance of BHyRL against baseline methods. Finally, we zero-transfer our learned 6D fetching policy with BHyRL to our MM robot TIAGo++. For more details and code release, please refer to our project site: irosalab.com/rlmmbp

Recent years have seen the emergence of non-cooperative objects in space, like failed satellites and space junk. These objects are usually operated or collected by free-float dual-arm space manipulators. Thanks to eliminating the difficulties of modeling and manual parameter-tuning, reinforcement learning (RL) methods have shown a more promising sign in the trajectory planning of space manipulators. Although previous studies demonstrate their effectiveness, they cannot be applied in tracking dynamic targets with unknown rotation (non-cooperative objects). In this paper, we proposed a learning system for motion planning of free-float dual-arm space manipulator (FFDASM) towards non-cooperative objects. Specifically, our method consists of two modules. Module I realizes the multi-target trajectory planning for two end-effectors within a large target space. Next, Module II takes as input the point clouds of the non-cooperative object to estimate the motional property, and then can predict the position of target points on an non-cooperative object. We leveraged the combination of Module I and Module II to track target points on a spinning object with unknown regularity successfully. Furthermore, the experiments also demonstrate the scalability and generalization of our learning system.

Decision-making is critical for lane change in autonomous driving. Reinforcement learning (RL) algorithms aim to identify the values of behaviors in various situations and thus they become a promising pathway to address the decision-making problem. However, poor runtime safety hinders RL-based decision-making strategies from complex driving tasks in practice. To address this problem, human demonstrations are incorporated into the RL-based decision-making strategy in this paper. Decisions made by human subjects in a driving simulator are treated as safe demonstrations, which are stored into the replay buffer and then utilized to enhance the training process of RL. A complex lane change task in an off-ramp scenario is established to examine the performance of the developed strategy. Simulation results suggest that human demonstrations can effectively improve the safety of decisions of RL. And the proposed strategy surpasses other existing learning-based decision-making strategies with respect to multiple driving performances.

The 5th Multi-disciplinary Conference on Reinforcement Learning and Decision Making (RLDM) 2022 took place at Brown University from 8-11 June. The programme included invited and contributed talks, workshops, and poster sessions. The goal of RLDM is to provide a platform for communication among all researchers interested in learning and decision making over time to achieve a goal. Over the last few decades, reinforcement learning and decision making have been the focus of an incredible wealth of research spanning a wide variety of fields including psychology, artificial intelligence, machine learning, operations research, control theory, neuroscience, economics and ethology. The interdisciplinary sharing of ideas has been key to many developments in the field, and the meeting is characterized by the multidisciplinarity of the presenters and attendees.

Building artificial intelligence (AI) that aligns with human values is an unsolved problem. Here we developed a human-in-the-loop research pipeline called Democratic AI, in which reinforcement learning is used to design a social mechanism that humans prefer by majority. A large group of humans played an online investment game that involved deciding whether to keep a monetary endowment or to share it with others for collective benefit. Shared revenue was returned to players under two different redistribution mechanisms, one designed by the AI and the other by humans. The AI discovered a mechanism that redressed initial wealth imbalance, sanctioned free riders and successfully won the majority vote. By optimising for human preferences, Democratic AI offers a proof of concept for value-aligned policy innovation.

In multi-agent reinforcement learning systems, the actions of one agent can have a negative impact on the rewards of other agents. One way to combat this problem is to let agents trade their rewards amongst each other. Motivated by this, this work applies a trading approach to a simulated scheduling environment, where the agents are responsible for the assignment of incoming jobs to compute cores. In this environment, reinforcement learning agents learn to trade successfully. The agents can trade the usage right of computational cores to process high-priority, high-reward jobs faster than low-priority, low-reward jobs. However, due to combinatorial effects, the action and observation spaces of a simple reinforcement learning agent in this environment scale exponentially with key parameters of the problem size. However, the exponential scaling behavior can be transformed into a linear one if the agent is split into several independent sub-units. We further improve this distributed architecture using agent-internal parameter sharing. Moreover, it can be extended to set the exchange prices autonomously. We show that in our scheduling environment, the advantages of a distributed agent architecture clearly outweigh more aggregated approaches. We demonstrate that the distributed agent architecture becomes even more performant using agent-internal parameter sharing. Finally, we investigate how two different reward functions affect autonomous pricing and the corresponding scheduling.

Since 2016 federated learning (FL) has been an evolving topic of discussion in the artificial intelligence (AI) research community. Applications of FL led to the development and study of federated reinforcement learning (FRL). Few works exist on the topic of FRL applied to autonomous vehicle (AV) platoons. In addition, most FRL works choose a single aggregation method (usually weight or gradient aggregation). We explore FRL's effectiveness as a means to improve AV platooning by designing and implementing an FRL framework atop a custom AV platoon environment. The application of FRL in AV platooning is studied under two scenarios: (1) Inter-platoon FRL (Inter-FRL) where FRL is applied to AVs across different platoons; (2) Intra-platoon FRL (Intra-FRL) where FRL is applied to AVs within a single platoon. Both Inter-FRL and Intra-FRL are applied to a custom AV platooning environment using both gradient and weight aggregation to observe the performance effects FRL can have on AV platoons relative to an AV platooning environment trained without FRL. It is concluded that Intra-FRL using weight aggregation (Intra-FRLWA) provides the best performance for controlling an AV platoon. In addition, we found that weight aggregation in FRL for AV platooning provides increases in performance relative to gradient aggregation. Finally, a performance analysis is conducted for Intra-FRLWA versus a platooning environment without FRL for platoons of length 3, 4 and 5 vehicles. It is concluded that Intra-FRLWA largely out-performs the platooning environment that is trained without FRL.

Offline RL algorithms must account for the fact that the dataset they are provided may leave many facets of the environment unknown. The most common way to approach this challenge is to employ pessimistic or conservative methods, which avoid behaviors that are too dissimilar from those in the training dataset. However, relying exclusively on conservatism has drawbacks: performance is sensitive to the exact degree of conservatism, and conservative objectives can recover highly suboptimal policies. In this work, we propose that offline RL methods should instead be adaptive in the presence of uncertainty. We show that acting optimally in offline RL in a Bayesian sense involves solving an implicit POMDP. As a result, optimal policies for offline RL must be adaptive, depending not just on the current state but rather all the transitions seen so far during evaluation.We present a model-free algorithm for approximating this optimal adaptive policy, and demonstrate the efficacy of learning such adaptive policies in offline RL benchmarks.

Deep Reinforcement Learning (DRL) algorithms can scale to previously intractable problems. The automation of profit generation in the stock market is possible using DRL, by combining the financial assets price "prediction" step and the "allocation" step of the portfolio in one unified process to produce fully autonomous systems capable of interacting with their environment to make optimal decisions through trial and error. This work represents a DRL model to generate profitable trades in the stock market, effectively overcoming the limitations of supervised learning approaches. We formulate the trading problem as a Partially Observed Markov Decision Process (POMDP) model, considering the constraints imposed by the stock market, such as liquidity and transaction costs. We then solve the formulated POMDP problem using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm reporting a 2.68 Sharpe Ratio on unseen data set (test data). From the point of view of stock market forecasting and the intelligent decision-making mechanism, this paper demonstrates the superiority of DRL in financial markets over other types of machine learning and proves its credibility and advantages of strategic decision-making.

Reinforcement learning has witnessed recent applications to a variety of tasks in quantum programming. The underlying assumption is that those tasks could be modeled as Markov Decision Processes (MDPs). Here, we investigate the feasibility of this assumption by exploring its consequences for two fundamental tasks in quantum programming: state preparation and gate compilation. By forming discrete MDPs, focusing exclusively on the single-qubit case (both with and without noise), we solve for the optimal policy exactly through policy iteration. We find optimal paths that correspond to the shortest possible sequence of gates to prepare a state, or compile a gate, up to some target accuracy. As an example, we find sequences of $H$ and $T$ gates with length as small as $11$ producing $\sim 99\%$ fidelity for states of the form $(HT)^{n} |0\rangle$ with values as large as $n=10^{10}$. In the presence of gate noise, we demonstrate how the optimal policy adapts to the effects of noisy gates in order to achieve a higher state fidelity. Our work shows that one can meaningfully impose a discrete, stochastic and Markovian nature to a continuous, deterministic and non-Markovian quantum evolution, and provides theoretical insight into why reinforcement learning may be successfully used to find optimally short gate sequences in quantum programming.