State abstraction can give rise to models of environments that are both compressed and useful, thereby enabling efficient sequential decision making. In this work, we offer the first formalism and analysis of the trade-off between compression and performance made in the context of state abstraction for Apprenticeship Learning. We build on Rate-Distortion theory, the classic Blahut-Arimoto algorithm, and the Information Bottleneck method to develop an algorithm for computing state abstractions that approximate the optimal tradeoff between compression and performance. We illustrate the power of this algorithmic structure to offer insights into effective abstraction, compression, and reinforcement learning through a mixture of analysis, visuals, and experimentation.
Abstract-As intelligent robots become more prevalent, methods to make interaction with the robots more accessible are increasingly important. Communicating the tasks that a person wants the robot to carry out via natural language, and training the robot to ground the natural language through demonstration, are especially appealing approaches for interaction, since they do not require a technical background. However, existing approaches map natural language commands to robot command languages that directly express the sequence of actions the robot should execute. This sequence is often specific to a particular situation and does not generalize to new situations. To address this problem, we present a system that grounds natural language commands into reward functions using demonstrations of different natural language commands being carried out in the environment. Because language is grounded to reward functions, rather than explicit actions that the robot can perform, commands can be high-level, carried out in novel environments autonomously, and even transferred to other robots with different action spaces. We demonstrate that our learned model can be both generalized to novel environments and transferred to a robot with a different action space than the action space used during training.
Abstract-Humans can ground natural language commands to tasks at both abstract and fine-grained levels of specificity. For instance, a human forklift operator can be instructed to perform a high-level action, like "grab a pallet" or a low-level action like "tilt back a little bit." While robots are also capable of grounding language commands to tasks, previous methods implicitly assume that all commands and tasks reside at a single, fixed level of abstraction. Additionally, methods that do not use multiple levels of abstraction encounter inefficient planning and execution times as they solve tasks at a single level of abstraction with large, intractable state-action spaces closely resembling real world complexity. In this work, by grounding commands to all the tasks or subtasks available in a hierarchical planning framework, we arrive at a model capable of interpreting language at multiple levels of specificity ranging from coarse to more granular. We show that the accuracy of the grounding procedure is improved when simultaneously inferring the degree of abstraction in language used to communicate the task. Leveraging hierarchy also improves efficiency: our proposed approach enables a robot to respond to a command within one second on 90% of our tasks, while baselines take over twenty seconds on half the tasks. Finally, we demonstrate that a real, physical robot can ground commands at multiple levels of abstraction allowing it to efficiently plan different subtasks within the same planning hierarchy.
To widen their accessibility and increase their utility, intelligent agents must be able to learn complex behaviors as specified by (non-expert) human users. Moreover, they will need to learn these behaviors within a reasonable amount of time while efficiently leveraging the sparse feedback a human trainer is capable of providing. Recent work has shown that human feedback can be characterized as a critique of an agent's current behavior rather than as an alternative reward signal to be maximized, culminating in the COnvergent Actor-Critic by Humans (COACH) algorithm for making direct policy updates based on human feedback. Our work builds on COACH, moving to a setting where the agent's policy is represented by a deep neural network. We employ a series of modifications on top of the original COACH algorithm that are critical for successfully learning behaviors from high-dimensional observations, while also satisfying the constraint of obtaining reduced sample complexity. We demonstrate the effectiveness of our Deep COACH algorithm in the rich 3D world of Minecraft with an agent that learns to complete tasks by mapping from raw pixels to actions using only real-time human feedback in 10-15 minutes of interaction.
Abstract-Often times, natural language commands issued to robots not only specify a particular target configuration or goal state but also outline constraints on how the robot goes about its execution. That is, the path taken to achieving some goal state is given equal importance to the goal state itself. One example of this could be instructing a wheeled robot to "go to the living room but avoid the kitchen," in order to avoid scuffing the floor. This class of behaviors poses a serious obstacle to existing language understanding for robotics approaches that map to either action sequences or goal state representations. Due to the non-Markovian nature of the objective, approaches in the former category must map to potentially unbounded action sequences whereas approaches in the latter category would require folding the entirety of a robot's trajectory into a (traditionally Markovian) state representation, resulting in an intractable decision-making problem. To resolve this challenge, we use a recently introduced probabilistic variant of Linear Temporal Logic (LTL) as a goal specification language for a Markov Decision Process (MDP). While demonstrating that standard neural sequence-to-sequence learning models can successfully ground language to this semantic representation, we also provide analysis that highlights generalization to novel, unseen logical forms as an open problem for this class of model. We evaluate our system within two simulated robot domains as well as on a physical robot, demonstrating accurate language grounding alongside a significant expansion in the space of interpretable robot behaviors.
Robots operating alongside humans in diverse, stochastic environments must be able to accurately interpret natural language commands. These instructions often fall into one of two categories: those that specify a goal condition or target state, and those that specify explicit actions, or how to perform a given task. Recent approaches have used reward functions as a semantic representation of goal-based commands, which allows for the use of a state-of-the-art planner to find a policy for the given task. However, these reward functions cannot be directly used to represent action-oriented commands. We introduce a new hybrid approach, the Deep Recurrent Action-Goal Grounding Network (DRAGGN), for task grounding and execution that handles natural language from either category as input, and generalizes to unseen environments. Our robotsimulation results demonstrate that a system successfully interpreting both goaloriented and action-oriented task specifications brings us closer to robust natural language understanding for human-robot interaction.
State abstraction has been an essential tool for dramatically improving the sample efficiency of reinforcement-learning algorithms. Indeed, by exposing and accentuating various types of latent structure within the environment, different classes of state abstraction have enabled improved theoretical guarantees and empirical performance. When dealing with state abstractions that capture structure in the value function, however, a standard assumption is that the true abstraction has been supplied or unrealistically computed a priori, leaving open the question of how to efficiently uncover such latent structure while jointly seeking out optimal behavior. Taking inspiration from the bandit literature, we propose that an agent seeking out latent task structure must explicitly represent and maintain its uncertainty over that structure as part of its overall uncertainty about the environment. We introduce a practical algorithm for doing this using two posterior distributions over state abstractions and abstract-state values. In empirically validating our approach, we find that substantial performance gains lie in the multi-task setting where tasks share a common, low-dimensional representation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.