Reliability quantification of deep reinforcement learning (DRL)-based control is a significant challenge for the practical application of artificial intelligence (AI) in safety-critical systems. This study proposes a method for quantifying the reliability of DRL-based control. First, an existing method, random noise distillation, was applied to the reliability evaluation to clarify the issues to be solved. Second, a novel method for reliability quantification was proposed to solve these issues. The reliability is quantified using two neural networks: reference and evaluator. They have the same structure with the same initial parameters. The outputs of the two networks were the same before training. During training, the evaluator network parameters were updated to maximize the difference between the reference and evaluator networks for trained data. Thus, the reliability of the DRL-based control for a state can be evaluated based on the difference in output between the two networks. The proposed method was applied to DQN-based control as an example of a simple task, and its effectiveness was demonstrated. Finally, the proposed method was applied to the problem of switching trained models depending on the state. Con-sequently, the performance of the DRL-based control was improved by switching the trained models according to their reliability.
Understanding a \textit{reinforcement learning} policy, which guides state-to-action mappings to maximize rewards, necessitates an accompanying explanation for human comprehension. In this paper, we introduce a set of \textit{linear temporal logic} formulae designed to provide explanations for policies, and an algorithm for searching through those formulae for the one that best explains a given policy. Our focus is on explanations that elucidate both the ultimate objectives accomplished by the policy and the prerequisite conditions it upholds throughout its execution. The effectiveness of our proposed approach is illustrated through a simulated game of capture-the-flag and a car-parking environment,
Currently, reinforcement learning (RL), especially deep RL, has received more and more attention in the research area. However, the security of RL has been an obvious problem due to the attack manners becoming mature. In order to defend against such adversarial attacks, several practical approaches are developed, such as adversarial training, data filtering, etc. However, these methods are mostly based on empirical algorithms and experiments, without rigorous theoretical analysis of the robustness of the algorithms. In this paper, we develop an algorithm to certify the robustness of a given policy offline with random smoothing, which could be proven and conducted as efficiently as ones without random smoothing. Experiments on different environments confirm the correctness of our algorithm.
Current reinforcement learning algorithms struggle in sparse and complex environments, most notably in long-horizon manipulation tasks entailing a plethora of different sequences. In this work, we propose the Intrinsically Guided Exploration from Large Language Models (IGE-LLMs) framework. By leveraging LLMs as an assistive intrinsic reward, IGE-LLMs guides the exploratory process in reinforcement learning to address intricate long-horizon with sparse rewards robotic manipulation tasks. We evaluate our framework and related intrinsic learning methods in an environment challenged with exploration, and a complex robotic manipulation task challenged by both exploration and long-horizons. Results show IGE-LLMs (i) exhibit notably higher performance over related intrinsic methods and the direct use of LLMs in decision-making, (ii) can be combined and complement existing learning methods highlighting its modularity, (iii) are fairly insensitive to different intrinsic scaling parameters, and (iv) maintain robustness against increased levels of uncertainty and horizons.
This paper tackles the critical challenge of object navigation in autonomous navigation systems, particularly focusing on the problem of target approach and episode termination in environments with long optimal episode length in Deep Reinforcement Learning (DRL) based methods. While effective in environment exploration and object localization, conventional DRL methods often struggle with optimal path planning and termination recognition due to a lack of depth information. To overcome these limitations, we propose a novel approach, namely the Depth-Inference Termination Agent (DITA), which incorporates a supervised model called the Judge Model to implicitly infer object-wise depth and decide termination jointly with reinforcement learning. We train our judge model along with reinforcement learning in parallel and supervise the former efficiently by reward signal. Our evaluation shows the method is demonstrating superior performance, we achieve a 9.3% gain on success rate than our baseline method across all room types and gain 51.2% improvements on long episodes environment while maintaining slightly better Success Weighted by Path Length (SPL). Code and resources, visualization are available at: https://github.com/HuskyKingdom/DITA_acml2023
We present RLLTE: a long-term evolution, extremely modular, and open-source framework for reinforcement learning (RL) research and application. Beyond delivering top-notch algorithm implementations, RLLTE also serves as a toolkit for developing algorithms. More specifically, RLLTE decouples the RL algorithms completely from the exploitation-exploration perspective, providing a large number of components to accelerate algorithm development and evolution. In particular, RLLTE is the first RL framework to build a comprehensive ecosystem, which includes model training, evaluation, deployment, benchmark hub, and large language model (LLM)-empowered copilot. RLLTE is expected to set standards for RL engineering practice and be highly stimulative for industry and academia. Our documentation, examples, and source code are available at https://github.com/RLE-Foundation/rllte.
The streaming temporal action segmentation (STAS) task, a supplementary task of temporal action segmentation (TAS), has not received adequate attention in the field of video understanding. Existing TAS methods are constrained to offline scenarios due to their heavy reliance on multimodal features and complete contextual information. The STAS task requires the model to classify each frame of the entire untrimmed video sequence clip by clip in time, thereby extending the applicability of TAS methods to online scenarios. However, directly applying existing TAS methods to SATS tasks results in significantly poor segmentation outcomes. In this paper, we thoroughly analyze the fundamental differences between STAS tasks and TAS tasks, attributing the severe performance degradation when transferring models to model bias and optimization dilemmas. We introduce an end-to-end streaming video temporal action segmentation model with reinforcement learning (SVTAS-RL). The end-to-end modeling method mitigates the modeling bias introduced by the change in task nature and enhances the feasibility of online solutions. Reinforcement learning is utilized to alleviate the optimization dilemma. Through extensive experiments, the SVTAS-RL model significantly outperforms existing STAS models and achieves competitive performance to the state-of-the-art TAS model on multiple datasets under the same evaluation criteria, demonstrating notable advantages on the ultra-long video dataset EGTEA. Code is available at https://github.com/Thinksky5124/SVTAS.
In this paper, we propose a new approach called Adaptive Behavioral Costs in Reinforcement Learning (ABC-RL) for training a human-like agent with competitive strength. While deep reinforcement learning agents have recently achieved superhuman performance in various video games, some of these unconstrained agents may exhibit actions, such as shaking and spinning, that are not typically observed in human behavior, resulting in peculiar gameplay experiences. To behave like humans and retain similar performance, ABC-RL augments behavioral limitations as cost signals in reinforcement learning with dynamically adjusted weights. Unlike traditional constrained policy optimization, we propose a new formulation that minimizes the behavioral costs subject to a constraint of the value function. By leveraging the augmented Lagrangian, our approach is an approximation of the Lagrangian adjustment, which handles the trade-off between the performance and the human-like behavior. Through experiments conducted on 3D games in DMLab-30 and Unity ML-Agents Toolkit, we demonstrate that ABC-RL achieves the same performance level while significantly reducing instances of shaking and spinning. These findings underscore the effectiveness of our proposed approach in promoting more natural and human-like behavior during gameplay.
Robots and animals both experience the world through their bodies and senses. Their embodiment constrains their experiences, ensuring they unfold continuously in space and time. As a result, the experiences of embodied agents are intrinsically correlated. Correlations create fundamental challenges for machine learning, as most techniques rely on the assumption that data are independent and identically distributed. In reinforcement learning, where data are directly collected from an agent's sequential experiences, violations of this assumption are often unavoidable. Here, we derive a method that overcomes this issue by exploiting the statistical mechanics of ergodic processes, which we term maximum diffusion reinforcement learning. By decorrelating agent experiences, our approach provably enables single-shot learning in continuous deployments over the course of individual task attempts. Moreover, we prove our approach generalizes well-known maximum entropy techniques, and robustly exceeds state-of-the-art performance across popular benchmarks. Our results at the nexus of physics, learning, and control form a foundation for transparent and reliable decision-making in embodied reinforcement learning agents.
Zero-shot reinforcement learning (RL) promises to provide agents that can perform any task in an environment after an offline, reward-free pre-training phase. Methods leveraging successor measures and successor features have shown strong performance in this setting, but require access to large heterogenous datasets for pre-training which cannot be expected for most real problems. Here, we explore how the performance of zero-shot RL methods degrades when trained on small homogeneous datasets, and propose fixes inspired by conservatism, a well-established feature of performant single-task offline RL algorithms. We evaluate our proposals across various datasets, domains and tasks, and show that conservative zero-shot RL algorithms outperform their non-conservative counterparts on low quality datasets, and perform no worse on high quality datasets. Somewhat surprisingly, our proposals also outperform baselines that get to see the task during training. Our code is available via https://enjeeneer.io/projects/zero-shot-rl/ .
In the Centralized-Radio Access Network (C-RAN) architecture, functions can be placed in the central or distributed locations. This architecture can offer higher capacity and cost savings but also puts strict requirements on the fronthaul (FH). Adaptive FH compression schemes that adapt the compression amount to varying FH traffic are promising approaches to deal with stringent FH requirements. In this work, we design such a compression scheme using a model-free off policy deep reinforcement learning algorithm which accounts for FH latency and packet loss constraints. Furthermore, this algorithm is designed for model transparency and interpretability which is crucial for AI trustworthiness in performance critical domains. We show that our algorithm can successfully choose an appropriate compression scheme while satisfying the constraints and exhibits a roughly 70\% increase in FH utilization compared to a reference scheme.
Behavioral experiments on the trust game have shown that trust and trustworthiness are universal among human beings, contradicting the prediction by assuming \emph{Homo economicus} in orthodox Economics. This means some mechanism must be at work that favors their emergence. Most previous explanations however need to resort to some factors based upon imitative learning, a simple version of social learning. Here, we turn to the paradigm of reinforcement learning, where individuals update their strategies by evaluating the long-term return through accumulated experience. Specifically, we investigate the trust game with the Q-learning algorithm, where each participant is associated with two evolving Q-tables that guide one's decision making as trustor and trustee respectively. In the pairwise scenario, we reveal that high levels of trust and trustworthiness emerge when individuals appreciate both their historical experience and returns in the future. Mechanistically, the evolution of the Q-tables shows a crossover that resembles human's psychological changes. We also provide the phase diagram for the game parameters, where the boundary analysis is conducted. These findings are robust when the scenario is extended to a latticed population. Our results thus provide a natural explanation for the emergence of trust and trustworthiness without external factors involved. More importantly, the proposed paradigm shows the potential in deciphering many puzzles in human behaviors.
Majority of off-policy reinforcement learning algorithms use overestimation bias control techniques. Most of these techniques rooted in heuristics, primarily addressing the consequences of overestimation rather than its fundamental origins. In this work we present a novel approach to the bias correction, similar in spirit to Double Q-Learning. We propose using a policy in form of a mixture with two components. Each policy component is maximized and assessed by separate networks, which removes any basis for the overestimation bias. Our approach shows promising near-SOTA results on a small set of MuJoCo environments.
A sudden roadblock on highways due to many reasons such as road maintenance, accidents, and car repair is a common situation we encounter almost daily. Autonomous Vehicles (AVs) equipped with sensors that can acquire vehicle dynamics such as speed, acceleration, and location can make intelligent decisions to change lanes before reaching a roadblock. A number of literature studies have examined car-following models and lane-changing models. However, only a few studies proposed an integrated car-following and lane-changing model, which has the potential to model practical driving maneuvers. Hence, in this paper, we present an integrated car-following and lane-changing decision-control system based on Deep Reinforcement Learning (DRL) to address this issue. Specifically, we consider a scenario where sudden construction work will be carried out along a highway. We model the scenario as a Markov Decision Process (MDP) and employ the well-known DQN algorithm to train the RL agent to make the appropriate decision accordingly (i.e., either stay in the same lane or change lanes). To overcome the delay and computational requirement of DRL algorithms, we adopt an MEC-assisted architecture where the RL agents are trained on MEC servers. We utilize the highly reputable SUMO simulator and OPENAI GYM to evaluate the performance of the proposed model under two policies; {\epsilon}-greedy policy and Boltzmann policy. The results unequivocally demonstrate that the DQN agent trained using the {\epsilon}-greedy policy significantly outperforms the one trained with the Boltzmann policy.
In many massive IoT communication scenarios, the IoT devices require coverage from dynamic units that can move close to the IoT devices and reduce the uplink energy consumption. A robust solution is to deploy a large number of UAVs (UAV swarm) to provide coverage and a better line of sight (LoS) for the IoT network. However, the study of these massive IoT scenarios with a massive number of serving units leads to high dimensional problems with high complexity. In this paper, we apply multi-agent deep reinforcement learning to address the high-dimensional problem that results from deploying a swarm of UAVs to collect fresh information from IoT devices. The target is to minimize the overall age of information in the IoT network. The results reveal that both cooperative and partially cooperative multi-agent deep reinforcement learning approaches are able to outperform the high-complexity centralized deep reinforcement learning approach, which stands helpless in large-scale networks.
Robust header compression (ROHC), critically positioned between the network and the MAC layers, plays an important role in modern wireless communication systems for improving data efficiency. This work investigates bi-directional ROHC (BD-ROHC) integrated with a novel architecture of reinforcement learning (RL). We formulate a partially observable \emph{Markov} decision process (POMDP), in which agent is the compressor, and the environment consists of the decompressor, channel and header source. Our work adopts the well-known deep Q-network (DQN), which takes the history of actions and observations as inputs, and outputs the Q-values of corresponding actions. Compared with the ideal dynamic programming (DP) proposed in the existing works, our method is scalable to the state, action and observation spaces. In contrast, DP often suffers from formidable computational complexity when the number of states becomes large due to long decompressor feedback delay and complex channel models. In addition, our method does not require prior knowledge of the transition dynamics and accurate observation dependency of the model, which are often not available in many practical applications.
Personalized adaptive interventions offer the opportunity to increase patient benefits, however, there are challenges in their planning and implementation. Once implemented, it is an important question whether personalized adaptive interventions are indeed clinically more effective compared to a fixed gold standard intervention. In this paper, we present an innovative N-of-1 trial study design testing whether implementing a personalized intervention by an online reinforcement learning agent is feasible and effective. Throughout, we use a new study on physical exercise recommendations to reduce pain in endometriosis for illustration. We describe the design of a contextual bandit recommendation agent and evaluate the agent in simulation studies. The results show that, first, implementing a personalized intervention by an online reinforcement learning agent is feasible. Second, such adaptive interventions have the potential to improve patients' benefits even if only few observations are available. As one challenge, they add complexity to the design and implementation process. In order to quantify the expected benefit, data from previous interventional studies is required. We expect our approach to be transferable to other interventions and clinical interventions.
Reinforcement Learning (RL) allows learning non-trivial robot control laws purely from data. However, many successful applications of RL have relied on ad-hoc regularizations, such as hand-crafted curricula, to regularize the learning performance. In this paper, we pair a recent algorithm for automatically building curricula with RL on massively parallelized simulations to learn a tracking controller for a spherical pendulum on a robotic arm via RL. Through an improved optimization scheme that better respects the non-Euclidean task structure, we allow the method to reliably generate curricula of trajectories to be tracked, resulting in faster and more robust learning compared to an RL baseline that does not exploit this form of structured learning. The learned policy matches the performance of an optimal control baseline on the real system, demonstrating the potential of curriculum RL to jointly learn state estimation and control for non-linear tracking tasks.
Curriculum reinforcement learning (CRL) allows solving complex tasks by generating a tailored sequence of learning tasks, starting from easy ones and subsequently increasing their difficulty. Although the potential of curricula in RL has been clearly shown in various works, it is less clear how to generate them for a given learning environment, resulting in various methods aiming to automate this task. In this work, we focus on framing curricula as interpolations between task distributions, which has previously been shown to be a viable approach to CRL. Identifying key issues of existing methods, we frame the generation of a curriculum as a constrained optimal transport problem between task distributions. Benchmarks show that this way of curriculum generation can improve upon existing CRL methods, yielding high performance in various tasks with different characteristics.
Deployment in hazardous environments requires robots to understand the risks associated with their actions and movements to prevent accidents. Despite its importance, these risks are not explicitly modeled by currently deployed locomotion controllers for legged robots. In this work, we propose a risk sensitive locomotion training method employing distributional reinforcement learning to consider safety explicitly. Instead of relying on a value expectation, we estimate the complete value distribution to account for uncertainty in the robot's interaction with the environment. The value distribution is consumed by a risk metric to extract risk sensitive value estimates. These are integrated into Proximal Policy Optimization (PPO) to derive our method, Distributional Proximal Policy Optimization (DPPO). The risk preference, ranging from risk-averse to risk-seeking, can be controlled by a single parameter, which enables to adjust the robot's behavior dynamically. Importantly, our approach removes the need for additional reward function tuning to achieve risk sensitivity. We show emergent risk sensitive locomotion behavior in simulation and on the quadrupedal robot ANYmal. Videos of the experiments and code are available at https://sites.google.com/leggedrobotics.com/risk-aware-locomotion.