Fundamental Performance Limits for Sensor-Based Robot Control and Policy Learning

@article{Majumdar2022FundamentalPL,
  title={Fundamental Performance Limits for Sensor-Based Robot Control and Policy Learning},
  author={Anirudha Majumdar and Vincent Pacelli},
  journal={ArXiv},
  year={2022},
  volume={abs/2202.00129}
}
—Our goal is to develop theory and algorithms for establishing fundamental limits on performance for a given task imposed by a robot’s sensors. In order to achieve this, we define a quantity that captures the amount of task-relevant information provided by a sensor. Using a novel version of the generalized Fano inequality from information theory, we demonstrate that this quantity provides an upper bound on the highest achievable expected reward for one-step decision making tasks. We then extend… 
2 Citations

Figures from this paper

A general class of combinatorial filters that can be minimized efficiently

A new algorithm for constraint repair is introduced that applies to a large sub- class of filters, subsuming three distinct special cases for which the possibility of optimal minimization in polynomial time was known earlier.

Leveraging Distributional Bias For Reactive Collision Avoidance under Uncertainty: A Kernel Embedding Approach

An extensive empirical study is conducted to show that the proposed distribution matching approach for collision avoidance with previous non-parametric and Gaussian approximated methods of reactive collision avoidance can infer distributional bias from sample-level information.

References

SHOWING 1-10 OF 36 REFERENCES

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.

Derivations for Linear Algebra and Optimization

Much of this section was copied and paraphrased from Heath’s Scientific Computing. Anyways. Suppose we are looking for an orthogonal transformation that annihilates desired components of a given

How Are Learned Perception-Based Controllers Impacted by the Limits of Robust Control?

This work empirically studies two popular families of controllers: RL and system identification-based H∞ control, using visually estimated system state and shows that the fundamental limits of robust control have corresponding implications for the sample-efficiency and performance of learned perception-based controllers.

What Is Robotics? Why Do We Need It and How Can We Get It?

The new discipline needs a departmental home in the universities which it can justify both intellectually and by its capacity to attract new diverse populations inspired by the age old human fascination with robots.

pomdp_py: A Framework to Build and Solve POMDP Problems

Pomdp_py is a general purpose Partially Observable Markov Decision Process (POMDP) library written in Python and Cython that enabled the authors' torso-actuated robot to perform object search in 3D.

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.

Sensor Lattices: Structures for Comparing Information Feedback

  • S. LaValle
  • Mathematics
    2019 12th International Workshop on Robot Motion and Control (RoMoCo)
  • 2019
This paper addresses the sensing uncertainty associated with the many-to-one mapping from a physical state space onto a sensor observation space. By studying preimages of this mapping for each

On Variational Bounds of Mutual Information

This work introduces a continuum of lower bounds that encompasses previous bounds and flexibly trades off bias and variance and demonstrates the effectiveness of these new bounds for estimation and representation learning.

How Fast Is Too Fast? The Role of Perception Latency in High-Speed Sense and Avoid

This is the first theoretical work in which perception and actuation limitations are jointly considered to study the performance of a robotic platform in high-speed navigation.

Perception-Action Cycle: Models, Architectures, and Hardware

This book will provide a snapshot and a resume of the current state-of-the-art of the ongoing research avenues concerning the perception-reason-action cycle and provide an informational resource and methodology for anyone interested in constructing and developing models, algorithms and systems of autonomous machines empowered with cognitive capabilities.