Task-Driven Estimation and Control via Information Bottlenecks

@article{Pacelli2019TaskDrivenEA,
  title={Task-Driven Estimation and Control via Information Bottlenecks},
  author={Vincent Pacelli and Anirudha Majumdar},
  journal={2019 International Conference on Robotics and Automation (ICRA)},
  year={2019},
  pages={2061-2067}
}
Our goal is to develop a principled and general algorithmic framework for task-driven estimation and control for robotic systems. State-of-the-art approaches for controlling robotic systems typically rely heavily on accurately estimating the full state of the robot (e.g., a running robot might estimate joint angles and velocities, torso state, and position relative to a goal). However, full state representations are often excessively rich for the specific task at hand and can lead to… 

Figures from this paper

Learning Task-Driven Control Policies via Information Bottlenecks

A reinforcement learning approach to synthesizing task-driven control policies for robotic systems equipped with rich sensory modalities by deriving a policy gradient-style algorithm that constrains actions to only depend on task-relevant information.

Learning to Actively Reduce Memory Requirements for Robot Control Tasks

This work presents a reinforcement learning framework that leverages an implementation of the group LASSO regularization to synthesize policies that employ low-dimensional and task-centric memory representations and actively reduce memory requirements.

Task-relevant Representation Learning for Networked Robotic Perception

This paper presents an algorithm to learn task-relevant representations of sensory data that are co-designed with a pre-trained robotic perception model's ultimate objective that achieves high accuracy and robust generalization on diverse tasks including Mars terrain classification with low-power deep learning accelerators, neural motion planning, and environmental timeseries classification.

Robust Control Under Uncertainty via Bounded Rationality and Differential Privacy

The theory of differential privacy is used to design controllers with bounded sensitivity to errors in state estimates, and to bound the amount of state information used for control in order to impose decision-making under bounded rationality.

Information-theoretic Abstraction of Semantic Octree Models for Integrated Perception and Planning

An approach that enables autonomous robots to build and compress semantic environment representations from point-cloud data and allows robots to prioritize individual semantic classes when generating the compressed trees, so as to design multi-resolution representations that retain the relevant semantic information while simultaneously discarding unwanted semantic categories.

On Direct vs Indirect Data-Driven Predictive Control

By revealing the existence of two non-asymptotic regimes for the performance of direct and indirect data-driven predictive control designs, this study suggests that neither approach is invariably superior and that the choice of design must, in practice, be informed by the available dataset.

Co-Design of Communication and Machine Inference for Cloud Robotics

This paper presents an algorithm to learn task-relevant representations of sensory data that are codesigned with a pre-trained robotic perception model’s ultimate objective that achieves high accuracy and robust generalization on diverse tasks including Mars terrain classification with low-power deep learning accelerators, neural motion planning, and environmental timeseries classification.

Robot Encoder Server Decoder Server Task Model Train-Time Only DNN Local Compute

    References

    SHOWING 1-10 OF 46 REFERENCES

    Attention and anticipation in fast visual-inertial navigation

    • L. CarloneS. Karaman
    • Computer Science
      2017 IEEE International Conference on Robotics and Automation (ICRA)
    • 2017
    This papers develops a computational approach for visual attention in robots that enables accurate visual-inertial navigation while the state of the art fails to track robot's motion during aggressive maneuvers.

    A Separation Principle for Control in the Age of Deep Learning

    The problem of defining and inferring a state for a control system based on complex, high-dimensional, highly uncertain measurement streams, such as videos, is reviewed and ideas for an optimal representation are extended to the dynamic case.

    Efficient computation of optimal actions

    • E. Todorov
    • Computer Science
      Proceedings of the National Academy of Sciences
    • 2009
    This work proposes a more structured formulation that greatly simplifies the construction of optimal control laws in both discrete and continuous domains, and enables computations that were not possible before.

    Integrated perception and control at high speed

    A method for robust high-speed quadrotor navigation through unknown cluttered environments using integrated perception and control that forgoes maintaining a map in favor of using only instantaneous depth information in the local frame, providing robustness in the presence of signiőcant state estimate uncertainty.

    LQG Control With Minimum Directed Information: Semidefinite Programming Approach

    This work identifies the state-space realization of the optimal policy, which can be synthesized by an efficient algorithm based on semidefinite programming, and shows that the optimal control law is a linear–Gaussian randomized policy.

    Information theoretic MPC for model-based reinforcement learning

    An information theoretic model predictive control algorithm capable of handling complex cost criteria and general nonlinear dynamics and using multi-layer neural networks as dynamics models to solve model-based reinforcement learning tasks is introduced.

    Planning with Information-Processing Constraints and Model Uncertainty in Markov Decision Processes

    This work provides a generalized value iteration scheme together with a convergence proof that can derive a unified solution from a single generalized variational principle for Markov Decision Problems.

    Control and Sensing Co-design

    This paper presents the first scalable LQG co-design algorithms to compute near-optimal policies with provable sub-optimality guarantees, and shows that a separation principle holds, which partially decouples the design of sensing, estimation, and control.

    Optimal control: linear quadratic methods

    This augmented edition of a respected text teaches the reader how to use linear quadratic Gaussian methods effectively for the design of control systems. It explores linear optimal control theory

    Steps Towards a Theory of Visual Information: Active Perception, Signal-to-Symbol Conversion and the Interplay Between Sensing and Control

    It is shown that the "actionable information gap" between the two can be reduced by exercising control on the sensing process, and therefore, senging, control and information are inextricably tied.