Model-based Reinforcement Learning and Control have demonstrated great potential in various sequential decision making problem domains, including in robotics settings. However, real-world robotics systems often present challenges that limit the applicability of those methods. In particular, we note two problems that jointly happen in many industrial systems: 1) Irregular/asynchronous observations and actions and 2) Dramatic changes in environment dynamics from an episode to another (e.g. varying payload inertial properties). We propose a general framework that overcomes those difficulties by meta-learning adaptive dynamics models for continuous-time prediction and control. The proposed approach is task-agnostic and can be adapted to new tasks in a straight-forward manner. We present evaluations in two different robot simulations and on a real industrial robot.

In Reinforcement Learning (RL), agents aim at maximizing cumulative rewards in a given environment. During the learning process, RL agents face the dilemma of exploitation and exploration: leveraging existing knowledge to acquire rewards or seeking potentially higher ones. Using uncertainty as a guiding principle provides an active and effective approach to solving this dilemma and ensemble-based methods are one of the prominent avenues for quantifying uncertainty. Nevertheless, conventional ensemble-based uncertainty estimation lacks an explicit prior, deviating from Bayesian principles. Besides, this method requires diversity among members to generate less biased uncertainty estimation results. To address the above problems, previous research has incorporated random functions as priors. Building upon these foundational efforts, our work introduces an innovative approach with delicately designed prior NNs, which can incorporate maximal diversity in the initial value functions of RL. Our method has demonstrated superior performance compared with the random prior approaches in solving classic control problems and general exploration tasks, significantly improving sample efficiency.

Policy gradient lies at the core of deep reinforcement learning (RL) in continuous domains. Despite much success, it is often observed in practice that RL training with policy gradient can fail for many reasons, even on standard control problems with known solutions. We propose a framework for understanding one inherent limitation of the policy gradient approach: the optimization landscape in the policy space can be extremely non-smooth or fractal for certain classes of MDPs, such that there does not exist gradient to be estimated in the first place. We draw on techniques from chaos theory and non-smooth analysis, and analyze the maximal Lyapunov exponents and H\"older exponents of the policy optimization objectives. Moreover, we develop a practical method that can estimate the local smoothness of objective function from samples to identify when the training process has encountered fractal landscapes. We show experiments to illustrate how some failure cases of policy optimization can be explained by such fractal landscapes.

The pre-train and fine-tune paradigm in machine learning has had dramatic success in a wide range of domains because the use of existing data or pre-trained models on the internet enables quick and easy learning of new tasks. We aim to enable this paradigm in robotic reinforcement learning, allowing a robot to learn a new task with little human effort by leveraging data and models from the Internet. However, reinforcement learning often requires significant human effort in the form of manual reward specification or environment resets, even if the policy is pre-trained. We introduce RoboFuME, a reset-free fine-tuning system that pre-trains a multi-task manipulation policy from diverse datasets of prior experiences and self-improves online to learn a target task with minimal human intervention. Our insights are to utilize calibrated offline reinforcement learning techniques to ensure efficient online fine-tuning of a pre-trained policy in the presence of distribution shifts and leverage pre-trained vision language models (VLMs) to build a robust reward classifier for autonomously providing reward signals during the online fine-tuning process. In a diverse set of five real robot manipulation tasks, we show that our method can incorporate data from an existing robot dataset collected at a different institution and improve on a target task within as little as 3 hours of autonomous real-world experience. We also demonstrate in simulation experiments that our method outperforms prior works that use different RL algorithms or different approaches for predicting rewards. Project website: https://robofume.github.io

At present, robots typically require extensive training to successfully accomplish a single task. However, to truly enhance their usefulness in real-world scenarios, robots should possess the capability to perform multiple tasks effectively. To address this need, various multi-task reinforcement learning (RL) algorithms have been developed, including multi-task proximal policy optimization (PPO), multi-task trust region policy optimization (TRPO), and multi-task soft-actor critic (SAC). Nevertheless, these algorithms demonstrate optimal performance only when operating within an environment or observation space that exhibits a similar distribution. In reality, such conditions are often not the norm, as robots may encounter scenarios or observations that differ from those on which they were trained. Addressing this challenge, algorithms like Q-Weighted Adversarial Learning (QWALE) attempt to tackle the issue by training the base algorithm (generating prior data) solely for a particular task, rendering it unsuitable for generalization across tasks. So, the aim of this research project is to enable a robotic arm to successfully execute seven distinct tasks within the Meta World environment. To achieve this, a multi-task soft actor-critic (MT-SAC) is employed to train the robotic arm. Subsequently, the trained model will serve as a source of prior data for the single-life RL algorithm. The effectiveness of this MT-QWALE algorithm will be assessed by conducting tests on various target positions (novel positions). In the end, a comparison is provided between the trained MT-SAC and the MT-QWALE algorithm where the MT-QWALE performs better. An ablation study demonstrates that MT-QWALE successfully completes tasks with a slightly larger number of steps even after hiding the final goal position.

Autonomous vehicles have the potential to revolutionize transportation, but they must be able to navigate safely in traffic before they can be deployed on public roads. The goal of this project is to train autonomous vehicles to make decisions to navigate in uncertain environments using deep reinforcement learning techniques using the CARLA simulator. The simulator provides a realistic and urban environment for training and testing self-driving models. Deep Q-Networks (DQN) are used to predict driving actions. The study involves the integration of collision sensors, segmentation, and depth camera for better object detection and distance estimation. The model is tested on 4 different trajectories in presence of different types of 4-wheeled vehicles and pedestrians. The segmentation and depth cameras were utilized to ensure accurate localization of objects and distance measurement. Our proposed method successfully navigated the self-driving vehicle to its final destination with a high success rate without colliding with other vehicles, pedestrians, or going on the sidewalk. To ensure the optimal performance of our reinforcement learning (RL) models in navigating complex traffic scenarios, we implemented a pre-processing step to reduce the state space. This involved processing the images and sensor output before feeding them into the model. Despite significantly decreasing the state space, our approach yielded robust models that successfully navigated through traffic with high levels of safety and accuracy.

Reinforcement learning (RL) has emerged as a powerful approach for tackling complex medical decision-making problems such as treatment planning, personalized medicine, and optimizing the scheduling of surgeries and appointments. It has gained significant attention in the field of Natural Language Processing (NLP) due to its ability to learn optimal strategies for tasks such as dialogue systems, machine translation, and question-answering. This paper presents a review of the RL techniques in NLP, highlighting key advancements, challenges, and applications in healthcare. The review begins by visualizing a roadmap of machine learning and its applications in healthcare. And then it explores the integration of RL with NLP tasks. We examined dialogue systems where RL enables the learning of conversational strategies, RL-based machine translation models, question-answering systems, text summarization, and information extraction. Additionally, ethical considerations and biases in RL-NLP systems are addressed.

Recent statistical and reinforcement learning methods have significantly advanced patient care strategies. However, these approaches face substantial challenges in high-stakes contexts, including missing data, inherent stochasticity, and the critical requirements for interpretability and patient safety. Our work operationalizes a safe and interpretable framework to identify optimal treatment regimes. This approach involves matching patients with similar medical and pharmacological characteristics, allowing us to construct an optimal policy via interpolation. We perform a comprehensive simulation study to demonstrate the framework's ability to identify optimal policies even in complex settings. Ultimately, we operationalize our approach to study regimes for treating seizures in critically ill patients. Our findings strongly support personalized treatment strategies based on a patient's medical history and pharmacological features. Notably, we identify that reducing medication doses for patients with mild and brief seizure episodes while adopting aggressive treatment for patients in intensive care unit experiencing intense seizures leads to more favorable outcomes.

Specifying objective functions for machine learning systems is challenging, and misspecified objectives can be hacked [1, 2] or incentivise degenerate behavior [3, 4, 5]. Techniques such as reinforcement learning from human feedback (RLHF) enable ML systems to instead learn appropriate objectives from human feedback [6, 7, 8]. These techniques are widely used to finetune large language models [9, 10, 11, 12] and to train reinforcement learning agents to perform complex maneuvers in continuous control environments [6, 7]. However, while RLHF is relied upon to ensure that these systems are safe, helpful, and harmless [13], it still faces many limitations and unsolved challenges [14]. In particular, RLHF systems typically rely on the assumption that all feedback comes from a single human teacher, despite gathering feedback from a range of teachers with varying levels of rationality and expertise. For example, Stiennon et al. [8], Bai et al. [13] and Ouyang et al. [15] assume that all feedback comes from a single teacher, but find that annotators and researchers actually disagree 23% to 37% of the time. Reward learning has been shown to be highly sensitive to incorrect assumptions about the process that generates feedback [16, 17, 18, 19], so this single-teacher assumption exposes these systems to dangerous failures [20]. Ideally, RLHF systems should consider the differences between each teacher to improve their safety and reliability. To leverage multiple teachers in RLHF, we introduce a novel problem called a Hidden Utility Bandit (HUB), illustrated in Figure 1.

The primacy bias in deep reinforcement learning (DRL), which refers to the agent's tendency to overfit early data and lose the ability to learn from new data, can significantly decrease the performance of DRL algorithms. Previous studies have shown that employing simple techniques, such as resetting the agent's parameters, can substantially alleviate the primacy bias. However, we observe that resetting the agent's parameters harms its performance in the context of model-based reinforcement learning (MBRL). In fact, on further investigation, we find that the primacy bias in MBRL differs from that in model-free RL. In this work, we focus on investigating the primacy bias in MBRL and propose world model resetting, which works in MBRL. We apply our method to two different MBRL algorithms, MBPO and DreamerV2. We validate the effectiveness of our method on multiple continuous control tasks on MuJoCo and DeepMind Control Suite, as well as discrete control tasks on Atari 100k benchmark. The results show that world model resetting can significantly alleviate the primacy bias in model-based setting and improve algorithm's performance. We also give a guide on how to perform world model resetting effectively.