Reinforcement Learning with Human Feedback (RLHF) is a paradigm in which an RL agent learns to optimize a task using pair-wise preference-based feedback over trajectories, rather than explicit reward signals. While RLHF has demonstrated practical success in fine-tuning language models, existing empirical work does not address the challenge of how to efficiently sample trajectory pairs for querying human feedback. In this study, we propose an efficient sampling approach to acquiring exploratory trajectories that enable accurate learning of hidden reward functions before collecting any human feedback. Theoretical analysis demonstrates that our algorithm requires less human feedback for learning the optimal policy under preference-based models with linear parameterization and unknown transitions, compared to the existing literature. Specifically, our framework can incorporate linear and low-rank MDPs. Additionally, we investigate RLHF with action-based comparison feedback and introduce an efficient querying algorithm tailored to this scenario.
相關內容
Procedural content generation (PCG) is a growing field, with numerous applications in the video game industry and great potential to help create better games at a fraction of the cost of manual creation. However, much of the work in PCG is focused on generating relatively straightforward levels in simple games, as it is challenging to design an optimisable objective function for complex settings. This limits the applicability of PCG to more complex and modern titles, hindering its adoption in industry. Our work aims to address this limitation by introducing a compositional level generation method that recursively composes simple low-level generators to construct large and complex creations. This approach allows for easily-optimisable objectives and the ability to design a complex structure in an interpretable way by referencing lower-level components. We empirically demonstrate that our method outperforms a non-compositional baseline by more accurately satisfying a designer's functional requirements in several tasks. Finally, we provide a qualitative showcase (in Minecraft) illustrating the large and complex, but still coherent, structures that were generated using simple base generators.
The current capabilities of robotic systems make human collaboration necessary to accomplish complex tasks effectively. In this work, we are introducing a framework to ensure safety in a human-robot collaborative environment. The system is composed of a wearable 2-DOF robot, a low-cost and easy-to-install tracking system, and a collision avoidance algorithm based on the Artificial Potential Field (APF). The wearable robot is designed to hold a fiducial marker and maintain its visibility to the tracking system, which, in turn, localizes the user's hand with good accuracy and low latency and provides haptic feedback to the user. The system is designed to enhance the performance of collaborative tasks while ensuring user safety. Three experiments were carried out to evaluate the performance of the proposed system. The first one evaluated the accuracy of the tracking system. The second experiment analyzed human-robot behavior during an imminent collision. The third experiment evaluated the system in a collaborative activity in a shared working environment. The results show that the implementation of the introduced system reduces the operation time by 16% and increases the average distance between the user's hand and the robot by 5 cm.
To enable safe and effective human-robot collaboration (HRC) in smart manufacturing, seamless integration of sensing, cognition, and prediction into the robot controller is critical for real-time awareness, response, and communication inside a heterogeneous environment (robots, humans, and equipment). The proposed approach takes advantage of the prediction capabilities of nonlinear model predictive control (NMPC) to execute a safe path planning based on feedback from a vision system. In order to satisfy the requirement of real-time path planning, an embedded solver based on a penalty method is applied. However, due to tight sampling times NMPC solutions are approximate, and hence the safety of the system cannot be guaranteed. To address this we formulate a novel safety-critical paradigm with an exponential control barrier function (ECBF) used as a safety filter. We also design a simple human-robot collaboration scenario using V-REP to evaluate the performance of the proposed controller and investigate whether integrating human pose prediction can help with safe and efficient collaboration. The robot uses OptiTrack cameras for perception and dynamically generates collision-free trajectories to the predicted target interactive position. Results for a number of different configurations confirm the efficiency of the proposed motion planning and execution framework. It yields a 19.8% reduction in execution time for the HRC task considered.
Robot swarms often exhibit emergent behaviors that are fascinating to observe; however, it is often difficult to predict what swarm behaviors can emerge under a given set of agent capabilities. We seek to efficiently leverage human input to automatically discover a taxonomy of collective behaviors that can emerge from a particular multi-agent system, without requiring the human to know beforehand what behaviors are interesting or even possible. Our proposed approach adapts to user preferences by learning a similarity space over swarm collective behaviors using self-supervised learning and human-in-the-loop queries. We combine our learned similarity metric with novelty search and clustering to explore and categorize the space of possible swarm behaviors. We also propose several general-purpose heuristics that improve the efficiency of our novelty search by prioritizing robot controllers that are likely to lead to interesting emergent behaviors. We test our approach in simulation on two robot capability models and show that our methods consistently discover a richer set of emergent behaviors than prior work. Code, videos, and datasets are available at //sites.google.com/view/evolving-novel-swarms.
Language models (LMs) have demonstrated their capability in possessing commonsense knowledge of the physical world, a crucial aspect of performing tasks in everyday life. However, it remains unclear **whether LMs have the capacity to generate grounded, executable plans for embodied tasks.** This is a challenging task as LMs lack the ability to perceive the environment through vision and feedback from the physical environment. In this paper, we address this important research question and present the first investigation into the topic. Our novel problem formulation, named **G-PlanET**, inputs a high-level goal and a data table about objects in a specific environment, and then outputs a step-by-step actionable plan for a robotic agent to follow. To facilitate the study, we establish an **evaluation protocol** and design a dedicated metric to assess the quality of the plans. Our experiments demonstrate that the use of tables for encoding the environment and an iterative decoding strategy can significantly enhance the LMs' ability in grounded planning. Our analysis also reveals interesting and non-trivial findings.
We evaluate benchmark deep reinforcement learning (DRL) algorithms on the task of portfolio optimisation under a simulator. The simulator is based on correlated geometric Brownian motion (GBM) with the Bertsimas-Lo (BL) market impact model. Using the Kelly criterion (log utility) as the objective, we can analytically derive the optimal policy without market impact and use it as an upper bound to measure performance when including market impact. We found that the off-policy algorithms DDPG, TD3 and SAC were unable to learn the right Q function due to the noisy rewards and therefore perform poorly. The on-policy algorithms PPO and A2C, with the use of generalised advantage estimation (GAE), were able to deal with the noise and derive a close to optimal policy. The clipping variant of PPO was found to be important in preventing the policy from deviating from the optimal once converged. In a more challenging environment where we have regime changes in the GBM parameters, we found that PPO, combined with a hidden Markov model (HMM) to learn and predict the regime context, is able to learn different policies adapted to each regime. Overall, we find that the sample complexity of these algorithms is too high, requiring more than 2m steps to learn a good policy in the simplest setting, which is equivalent to almost 8,000 years of daily prices.
This paper studies online convex optimization with stochastic constraints. We propose a variant of the drift-plus-penalty algorithm that guarantees $O(\sqrt{T})$ expected regret and zero constraint violation, after a fixed number of iterations, which improves the vanilla drift-plus-penalty method with $O(\sqrt{T})$ constraint violation. Our algorithm is oblivious to the length of the time horizon $T$, in contrast to the vanilla drift-plus-penalty method. This is based on our novel drift lemma that provides time-varying bounds on the virtual queue drift and, as a result, leads to time-varying bounds on the expected virtual queue length. Moreover, we extend our framework to stochastic-constrained online convex optimization under two-point bandit feedback. We show that by adapting our algorithmic framework to the bandit feedback setting, we may still achieve $O(\sqrt{T})$ expected regret and zero constraint violation, improving upon the previous work for the case of identical constraint functions. Numerical results demonstrate our theoretical results.
We consider a persuasion problem between a sender and a receiver whose utility may be nonlinear in her belief; we call such receivers risk-conscious. Such utility models arise when the receiver exhibits systematic biases away from expected-utility-maximization, such as uncertainty aversion (e.g., from sensitivity to the variance of the waiting time for a service). Due to this nonlinearity, the standard approach to finding the optimal persuasion mechanism using revelation principle fails. To overcome this difficulty, we use the underlying geometry of the problem to develop a convex optimization framework to find the optimal persuasion mechanism. We define the notion of full persuasion and use our framework to characterize conditions under which full persuasion can be achieved. We use our approach to study binary persuasion, where the receiver has two actions and the sender strictly prefers one of them at every state. Under a convexity assumption, we show that the binary persuasion problem reduces to a linear program, and establish a canonical set of signals where each signal either reveals the state or induces in the receiver uncertainty between two states. Finally, we discuss the broader applicability of our methods to more general contexts, and illustrate our methodology by studying information sharing of waiting times in service systems.
Policies often fail due to distribution shift -- changes in the state and reward that occur when a policy is deployed in new environments. Data augmentation can increase robustness by making the model invariant to task-irrelevant changes in the agent's observation. However, designers don't know which concepts are irrelevant a priori, especially when different end users have different preferences about how the task is performed. We propose an interactive framework to leverage feedback directly from the user to identify personalized task-irrelevant concepts. Our key idea is to generate counterfactual demonstrations that allow users to quickly identify possible task-relevant and irrelevant concepts. The knowledge of task-irrelevant concepts is then used to perform data augmentation and thus obtain a policy adapted to personalized user objectives. We present experiments validating our framework on discrete and continuous control tasks with real human users. Our method (1) enables users to better understand agent failure, (2) reduces the number of demonstrations required for fine-tuning, and (3) aligns the agent to individual user task preferences.
In large-scale systems there are fundamental challenges when centralised techniques are used for task allocation. The number of interactions is limited by resource constraints such as on computation, storage, and network communication. We can increase scalability by implementing the system as a distributed task-allocation system, sharing tasks across many agents. However, this also increases the resource cost of communications and synchronisation, and is difficult to scale. In this paper we present four algorithms to solve these problems. The combination of these algorithms enable each agent to improve their task allocation strategy through reinforcement learning, while changing how much they explore the system in response to how optimal they believe their current strategy is, given their past experience. We focus on distributed agent systems where the agents' behaviours are constrained by resource usage limits, limiting agents to local rather than system-wide knowledge. We evaluate these algorithms in a simulated environment where agents are given a task composed of multiple subtasks that must be allocated to other agents with differing capabilities, to then carry out those tasks. We also simulate real-life system effects such as networking instability. Our solution is shown to solve the task allocation problem to 6.7% of the theoretical optimal within the system configurations considered. It provides 5x better performance recovery over no-knowledge retention approaches when system connectivity is impacted, and is tested against systems up to 100 agents with less than a 9% impact on the algorithms' performance.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191