Robust Control Under Uncertainty via Bounded Rationality and Differential Privacy

  title={Robust Control Under Uncertainty via Bounded Rationality and Differential Privacy},
  author={Vincent Pacelli and Anirudha Majumdar},
  journal={2022 International Conference on Robotics and Automation (ICRA)},
The rapid development of affordable and compact high-fidelity sensors (e.g., cameras and LIDAR) allows robots to construct detailed estimates of their states and environments. However, the availability of such rich sensor information introduces two challenges: (i) the lack of analytic sensing models, which makes it difficult to design controllers that are robust to sensor failures, and (ii) the computational expense of processing the high-dimensional sensor information in real time. This paper… 

Figures from this paper

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

A novel version of the generalized Fano inequality from information theory is used to demonstrate that a quantity that captures the amount of task-relevant information provided by a sensor provides an upper bound on the highest achievable expected reward for one-step decision making tasks.

Differentially Private Timeseries Forecasts for Networked Control

This work analyzes a scenario where the controller is able to pay the forecasting models an economic incentive, such as money, to lower their noise, and solves a biconvex optimization problem with guaranteed local optimality.

Decision-Making Among Bounded Rational Agents

Bounded rationality directly models the agent’s information processing ability, which is represented as the KL-divergence between nominal and optimized stochastic policies, and the solution to the bounded-optimal policy can be obtained by an e-cient importance sampling approach.



Differentially Private Filtering

Methods are developed to approximate a given filter by a differentially private version, so that the distortion introduced by the privacy mechanism is minimized, and extended to dynamic systems with many participants contributing independent input signals.

Task-Driven Estimation and Control via Information Bottlenecks

This work proposes novel iterative algorithms for automatically synthesizing a task-driven representation (given in terms of a set of task-relevant variables (TRVs) and a performant control policy that is a function of the TRVs and presents online algorithms for estimating the TRV in order to apply the control policy.

Differential privacy in control and network systems

This tutorial paper gives a systems and control perspective on the topic of privacy preserving data analysis, with a particular emphasis on the processing of dynamic data as well as data exchanged in networks.

Information-Theoretic Bounded Rationality

This paper provides a conceptual justification for using the free energy functional as the objective function for characterizing bounded-rational decisions and discusses the single-step decision-making case, and shows how to extend it to sequential decisions using equivalence transformations.

Dual Online Stein Variational Inference for Control and Dynamics

An implicit variational inference algorithm able to estimate distributions over model parameters and control inputs on-the-fly on the fly is devised, enabling the approximation of complex multi-modal posterior distributions, typically occurring in challenging and realistic robot navigation tasks.

Differential Privacy for Bayesian Inference through Posterior Sampling

This work defines differential privacy over arbitrary data set metrics, outcome spaces and distribution families, and proves bounds on the sensitivity of the posterior to the data, which delivers a measure of robustness.

Finite-time regional verification of stochastic non-linear systems

This work considers the problem of bounding the probability of failure over a finite time for stochastic non-linear systems with continuous state, and searches for exponential barrier functions that provide bounds using a variant of the classical supermartingale result.

Privacy in Control and Dynamical Systems

This review focuses on differential privacy use in the context of control systems, in which the data under processing often take the form of data streams, and illustrates the use of mathematical tools from control and optimization to convert a nonprivate algorithm to its private counterpart.

An Adversarial Interpretation of Information-Theoretic Bounded Rationality

It turns out that the optimal strategy of the adversary consists in choosing costs so as to render the decision maker indifferent among its choices, which is a definining property of a Nash equilibrium, thus tightening the connection between free energy optimization and game theory.

Differential privacy for dynamical sensitive data

This work protects dynamical systems from being tracked by an adversary by providing differentially private guarantees by proposing a mechanism which adds artificial noise to the input of the system and the measurements which are then published.