Deep reinforcement learning (DRL) has achieved tremendous success in many complex decision-making tasks of autonomous systems with high-dimensional state and/or action spaces. However, the safety and stability still remain major concerns that hinder the applications of DRL to safety-critical autonomous systems. To address the concerns, we proposed the Phy-DRL: a physical deep reinforcement learning framework. The Phy-DRL is novel in two architectural designs: i) Lyapunov-like reward, and ii) residual control (i.e., integration of physics-model-based control and data-driven control). The concurrent physical reward and residual control empower the Phy-DRL the (mathematically) provable safety and stability guarantees. Through experiments on the inverted pendulum, we show that the Phy-DRL features guaranteed safety and stability and enhanced robustness, while offering remarkably accelerated training and enlarged reward.

We introduce a novel deep reinforcement learning (DRL) approach to jointly optimize transmit beamforming and reconfigurable intelligent surface (RIS) phase shifts in a multiuser multiple input single output (MU-MISO) system to maximize the sum downlink rate under the phase-dependent reflection amplitude model. Our approach addresses the challenge of imperfect channel state information (CSI) and hardware impairments by considering a practical RIS amplitude model. We compare the performance of our approach against a vanilla DRL agent in two scenarios: perfect CSI and phase-dependent RIS amplitudes, and mismatched CSI and ideal RIS reflections. The results demonstrate that the proposed framework significantly outperforms the vanilla DRL agent under mismatch and approaches the golden standard. Our contributions include modifications to the DRL approach to address the joint design of transmit beamforming and phase shifts and the phase-dependent amplitude model. To the best of our knowledge, our method is the first DRL-based approach for the phase-dependent reflection amplitude model in RIS-aided MU-MISO systems. Our findings in this study highlight the potential of our approach as a promising solution to overcome hardware impairments in RIS-aided wireless communication systems.

Autonomous excavation is a challenging task. The unknown contact dynamics between the excavator bucket and the terrain could easily result in large contact forces and jamming problems during excavation. Traditional model-based methods struggle to handle such problems due to complex dynamic modeling. In this paper, we formulate the excavation skills with three novel manipulation primitives. We propose to learn the manipulation primitives with offline reinforcement learning (RL) to avoid large amounts of online robot interactions. The proposed method can learn efficient penetration skills from sub-optimal demonstrations, which contain sub-trajectories that can be ``stitched" together to formulate an optimal trajectory without causing jamming. We evaluate the proposed method with extensive experiments on excavating a variety of rigid objects and demonstrate that the learned policy outperforms the demonstrations. We also show that the learned policy can quickly adapt to unseen and challenging fragmented rocks with online fine-tuning.

Recent works have shown that sequence modeling can be effectively used to train reinforcement learning (RL) policies. However, the success of applying existing sequence models to planning, in which we wish to obtain a trajectory of actions to reach some goal, is less straightforward. The typical autoregressive generation procedures of sequence models preclude sequential refinement of earlier steps, which limits the effectiveness of a predicted plan. In this paper, we suggest an approach towards integrating planning with sequence models based on the idea of iterative energy minimization, and illustrate how such a procedure leads to improved RL performance across different tasks. We train a masked language model to capture an implicit energy function over trajectories of actions, and formulate planning as finding a trajectory of actions with minimum energy. We illustrate how this procedure enables improved performance over recent approaches across BabyAI and Atari environments. We further demonstrate unique benefits of our iterative optimization procedure, involving new task generalization, test-time constraints adaptation, and the ability to compose plans together. Project website: https://hychen-naza.github.io/projects/LEAP

How well do reward functions learned with inverse reinforcement learning (IRL) generalize? We illustrate that state-of-the-art IRL algorithms, which maximize a maximum-entropy objective, learn rewards that overfit to the demonstrations. Such rewards struggle to provide meaningful rewards for states not covered by the demonstrations, a major detriment when using the reward to learn policies in new situations. We introduce BC-IRL, a new inverse reinforcement learning method that learns reward functions that generalize better when compared to maximum-entropy IRL approaches. In contrast to the MaxEnt framework, which learns to maximize rewards around demonstrations, BC-IRL updates reward parameters such that the policy trained with the new reward matches the expert demonstrations better. We show that BC-IRL learns rewards that generalize better on an illustrative simple task and two continuous robotic control tasks, achieving over twice the success rate of baselines in challenging generalization settings. A visualization of learned rewards on a task where a 2D agent must navigate to the goal at the center. Figure 1a: Four trajectories are provided as demonstrations and the demonstrated states are visualized as points. Rewards learned via Maximum Entropy are in Figure 1b and BC-IRL in Figure 1c. Lighter colors represent larger predicted rewards. Reinforcement learning has demonstrated success on a broad range of tasks from navigation Wijmans et al. (2019), locomotion Kumar et al. (2021); Iscen et al. (2018), and manipulation Kalashnikov et al. (2018). However, this success depends on specifying an accurate and informative reward signal to guide the agent towards solving the task. For instance, imagine designing a reward function for a robot window cleaning task.

The FC Portugal 3D team is developed upon the structure of our previous Simulation league 2D/3D teams and our standard platform league team. Our research concerning the robot low-level skills is focused on developing behaviors that may be applied on real robots with minimal adaptation using model-based approaches. Our research on high-level soccer coordination methodologies and team playing is mainly focused on the adaptation of previously developed methodologies from our 2D soccer teams to the 3D humanoid environment and on creating new coordination methodologies based on the previously developed ones. The research-oriented development of our team has been pushing it to be one of the most competitive over the years (World champion in 2000 and Coach Champion in 2002, European champion in 2000 and 2001, Coach 2nd place in 2003 and 2004, European champion in Rescue Simulation and Simulation 3D in 2006, World Champion in Simulation 3D in Bremen 2006 and European champion in 2007, 2012, 2013, 2014 and 2015). This paper describes some of the main innovations of our 3D simulation league team during the last years. A new generic framework for reinforcement learning tasks has also been developed. The current research is focused on improving the above-mentioned framework by developing new learning algorithms to optimize low-level skills, such as running and sprinting. We are also trying to increase student contact by providing reinforcement learning assignments to be completed using our new framework, which exposes a simple interface without sharing low-level implementation details.

In this work, we study the concentration behavior of a stochastic approximation (SA) algorithm under a contractive operator with respect to an arbitrary norm. We consider two settings where the iterates are potentially unbounded: (1) bounded multiplicative noise, and (2) additive sub-Gaussian noise. We obtain maximal concentration inequalities on the convergence errors, and show that these errors have sub-Gaussian tails in the additive noise setting, and super-polynomial tails (faster than polynomial decay) in the multiplicative noise setting. In addition, we provide an impossibility result showing that it is in general not possible to achieve sub-exponential tails for SA with multiplicative noise. To establish these results, we develop a novel bootstrapping argument that involves bounding the moment generating function of the generalized Moreau envelope of the error and the construction of an exponential supermartingale to enable using Ville's maximal inequality. To demonstrate the applicability of our theoretical results, we use them to provide maximal concentration bounds for a large class of reinforcement learning algorithms, including but not limited to on-policy TD-learning with linear function approximation, off-policy TD-learning with generalized importance sampling factors, and $Q$-learning. To the best of our knowledge, super-polynomial concentration bounds for off-policy TD-learning have not been established in the literature due to the challenge of handling the combination of unbounded iterates and multiplicative noise.

Existing learning methods for LiDAR-based applications use 3D points scanned under a pre-determined beam configuration, e.g., the elevation angles of beams are often evenly distributed. Those fixed configurations are task-agnostic, so simply using them can lead to sub-optimal performance. In this work, we take a new route to learn to optimize the LiDAR beam configuration for a given application. Specifically, we propose a reinforcement learning-based learning-to-optimize (RL-L2O) framework to automatically optimize the beam configuration in an end-to-end manner for different LiDAR-based applications. The optimization is guided by the final performance of the target task and thus our method can be integrated easily with any LiDAR-based application as a simple drop-in module. The method is especially useful when a low-resolution (low-cost) LiDAR is needed, for instance, for system deployment at a massive scale. We use our method to search for the beam configuration of a low-resolution LiDAR for two important tasks: 3D object detection and localization. Experiments show that the proposed RL-L2O method improves the performance in both tasks significantly compared to the baseline methods. We believe that a combination of our method with the recent advances of programmable LiDARs can start a new research direction for LiDAR-based active perception. The code is publicly available at https://github.com/vniclas/lidar_beam_selection

Most offline reinforcement learning (RL) methods suffer from the trade-off between improving the policy to surpass the behavior policy and constraining the policy to limit the deviation from the behavior policy as computing $Q$-values using out-of-distribution (OOD) actions will suffer from errors due to distributional shift. The recently proposed \textit{In-sample Learning} paradigm (i.e., IQL), which improves the policy by quantile regression using only data samples, shows great promise because it learns an optimal policy without querying the value function of any unseen actions. However, it remains unclear how this type of method handles the distributional shift in learning the value function. In this work, we make a key finding that the in-sample learning paradigm arises under the \textit{Implicit Value Regularization} (IVR) framework. This gives a deeper understanding of why the in-sample learning paradigm works, i.e., it applies implicit value regularization to the policy. Based on the IVR framework, we further propose two practical algorithms, Sparse $Q$-learning (SQL) and Exponential $Q$-learning (EQL), which adopt the same value regularization used in existing works, but in a complete in-sample manner. Compared with IQL, we find that our algorithms introduce sparsity in learning the value function, making them more robust in noisy data regimes. We also verify the effectiveness of SQL and EQL on D4RL benchmark datasets and show the benefits of in-sample learning by comparing them with CQL in small data regimes.

Applying reinforcement learning (RL) to sparse reward domains is notoriously challenging due to insufficient guiding signals. Common RL techniques for addressing such domains include (1) learning from demonstrations and (2) curriculum learning. While these two approaches have been studied in detail, they have rarely been considered together. This paper aims to do so by introducing a principled task phasing approach that uses demonstrations to automatically generate a curriculum sequence. Using inverse RL from (suboptimal) demonstrations we define a simple initial task. Our task phasing approach then provides a framework to gradually increase the complexity of the task all the way to the target task, while retuning the RL agent in each phasing iteration. Two approaches for phasing are considered: (1) gradually increasing the proportion of time steps an RL agent is in control, and (2) phasing out a guiding informative reward function. We present conditions that guarantee the convergence of these approaches to an optimal policy. Experimental results on 3 sparse reward domains demonstrate that our task phasing approaches outperform state-of-the-art approaches with respect to asymptotic performance.