Annual Review of Control, Robotics, and Autonomous Systems - Volume 4, 2021

Robotics is an emerging synthetic science concerned with programming work. Robot technologies are quickly advancing beyond the insights of the existing science. More secure intellectual foundations will be required to achieve better, more reliable, and safer capabilities as their penetration into society deepens. Presently missing foundations include the identification of fundamental physical limits, the development of new dynamical systems theory, and the invention of physically grounded programming languages. 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.

Physics-based simulation provides an accelerated and safe avenue for developing, verifying, and testing robotic control algorithms and prototype designs. In the quest to leverage machine learning for developing AI-enabled robots, physics-based simulation can generate a wealth of labeled training data in a short amount of time. Physics-based simulation also creates an ideal proving ground for developing intelligent robots that can both learn from their mistakes and be verifiable. This article provides an overview of the use of simulation in robotics, emphasizing how robots (with sensing and actuation components), the environment they operate in, and the humans they interact with are simulated in practice. It concludes with an overview of existing tools for simulation in robotics and a short discussion of aspects that limit the role that simulation plays today in intelligent robot design.

A common way to represent a system's dynamics is to specify how the state evolves in time. An alternative viewpoint is to specify how functions of the state evolve in time. This evolution of functions is governed by a linear operator called the Koopman operator, whose spectral properties reveal intrinsic features of a system. For instance, its eigenfunctions determine coordinates in which the dynamics evolve linearly. This review discusses the theoretical foundations of Koopman operator methods, as well as numerical methods developed over the past two decades to approximate the Koopman operator from data, for systems both with and without actuation. We pay special attention to ergodic systems, for which especially effective numerical methods are available. For nonlinear systems with an affine control input, the Koopman formalism leads naturally to systems that are bilinear in the state and the input, and this structure can be leveraged for the design of controllers and estimators.

Optimal transport began as the problem of how to efficiently redistribute goods between production and consumers and evolved into a far-reaching geometric variational framework for studying flows of distributions on metric spaces. This theory enables a class of stochastic control problems to regulate dynamical systems so as to limit uncertainty to within specified limits. Representative control examples include the landing of a spacecraft aimed probabilistically toward a target and the suppression of undesirable effects of thermal noise on resonators; in both of these examples, the goal is to regulate the flow of the distribution of the random state. A most unlikely link turned up between transport of probability distributions and a maximum entropy inference problem posed by Erwin Schrödinger, where the latter is seen as an entropy-regularized version of the former. These intertwined topics of optimal transport, stochastic control, and inference are the subject of this review, which aims to highlight connections, insights, and computational tools while touching on quadratic regulator theory and probabilistic flows in discrete spaces and networks.

In this review, we present a comprehensive perspective on communication-aware robotics, an area that considers realistic communication environments and aims to jointly optimize communication and navigation. The main focus of the article is theoretical characterization and understanding of performance guarantees. We begin by summarizing the best prediction an unmanned vehicle can have of the channel quality at unvisited locations. We then consider the case of a single robot, showing how it can mathematically characterize the statistics of its traveled distance until connectivity and further plan its path to reach a connected location with optimality guarantees, in real channel environments and with minimum energy consumption. We then move to the case of multiple robots, showing how they can utilize their motions to enable robust information flow. We consider two specific robotic network configurations—robotic beamformers and robotic routers—and mathematically characterize properties of the co-optimum motion–communication decisions.

Many estimation, planning, and optimal control problems in robotics have an optimization problem at their core. In most of these optimization problems, the objective to be maximized or minimized is composed of many different factors or terms that are local in nature—that is, they depend only on a small subset of the variables. A particularly insightful way of modeling this locality structure is to use the concept of factor graphs, a bipartite graphical model in which factors represent functions on subsets of variables. Factor graphs can represent a wide variety of problems across robotics, expose opportunities to improve computational performance, and are beneficial in designing and thinking about how to model a problem, even aside from performance considerations. I discuss each of these three aspects in detail and review several state-of-the-art robotics applications in which factor graphs have been used with great success.

Brain–machine interfaces (BMIs) promise to restore movement and communication in people with paralysis and ultimately allow the human brain to interact seamlessly with external devices, paving the way for a new wave of medical and consumer technology. However, neural activity can adapt and change over time, presenting a substantial challenge for reliable BMI implementation. Large-scale recordings in animal studies now allow us to study how behavioral information is distributed in multiple brain areas, and state-of-the-art interfaces now incorporate models of the brain as a feedback controller. Ongoing research aims to understand the impact of neural plasticity on BMIs and find ways to leverage learning while accommodating unexpected changes in the neural code. We review the current state of experimental and clinical BMI research, focusing on what we know about the neural code, methods for optimizing decoders for closed-loop control, and emerging strategies for addressing neural plasticity.

The last decade has seen a flowering of applications driven by brain–machine interfaces (BMIs), particularly brain-actuated robotic devices designed to restore the independence of people suffering from severe motor disabilities. This review provides an overview of the state of the art of noninvasive BMI-driven devices based on 86 studies published in the last 15 years, with an emphasis on the interactions among the user, the BMI system, and the robot. We found that BMIs are used mostly to drive devices for navigation (e.g., telepresence mobile robots), with BMI paradigms based mainly on exogenous stimulation, and the majority of brain-actuated robots adopt a discrete control strategy. Most critically, in only a few works have disabled people evaluated a brain-actuated robot. The review highlights the most urgent challenges in the field, from the integration between BMI and robotics to the need for a user-centered design to boost the translational impact of BMIs.

Simultaneous localization and mapping (SLAM) is the process of constructing a global model of an environment from local observations of it; this is a foundational capability for mobile robots, supporting such core functions as planning, navigation, and control. This article reviews recent progress in SLAM, focusing on advances in the expressive capacity of the environmental models used in SLAM systems (representation) and the performance of the algorithms used to estimate these models from data (inference). A prominent theme of recent SLAM research is the pursuit of environmental representations (including learned representations) that go beyond the classical attributes of geometry and appearance to model properties such as hierarchical organization, affordance, dynamics, and semantics; these advances equip autonomous agents with a more comprehensive understanding of the world, enabling more versatile and intelligent operation. A second major theme is a revitalized interest in the mathematical properties of the SLAM estimation problem itself (including its computational and information-theoretic performance limits); this work has led to the development of novel classes of certifiable and robust inference methods that dramatically improve the reliability of SLAM systems in real-world operation. We survey these advances with an emphasis on their ramifications for achieving robust, long-duration autonomy, and conclude with a discussion of open challenges and a perspective on future research directions.

This article surveys recent advancements in strategy designs for persistent robotic surveillance tasks, with a focus on stochastic approaches. The problem describes how mobile robots stochastically patrol a graph in an efficient way, where the efficiency is defined with respect to relevant underlying performance metrics. We start by reviewing the basics of Markov chains, which are the primary motion models for stochastic robotic surveillance. We then discuss the two main criteria regarding the speed and unpredictability of surveillance strategies. The central objects that appear throughout the treatment are the hitting times of Markov chains, their distributions, and their expectations. We formulate various optimization problems based on the relevant metrics in different scenarios and establish their respective properties.

The problem of planning for a robot that operates in environments containing a large number of objects, taking actions to move itself through the world as well as to change the state of the objects, is known as task and motion planning (TAMP). TAMP problems contain elements of discrete task planning, discrete–continuous mathematical programming, and continuous motion planning and thus cannot be effectively addressed by any of these fields directly. In this article, we define a class of TAMP problems and survey algorithms for solving them, characterizing the solution methods in terms of their strategies for solving the continuous-space subproblems and their techniques for integrating the discrete and continuous components of the search.

Motion planning is a fundamental problem in autonomous robotics that requires finding a path to a specified goal that avoids obstacles and takes into account a robot's limitations and constraints. It is often desirable for this path to also optimize a cost function, such as path length. Formal path-quality guarantees for continuously valued search spaces are an active area of research interest. Recent results have proven that some sampling-based planning methods probabilistically converge toward the optimal solution as computational effort approaches infinity. This article summarizes the assumptions behind these popular asymptotically optimal techniques and provides an introduction to the significant ongoing research on this topic.

In this article, we first present some foundational results about the stability and positive stabilization of continuous-time positive systems. Necessary and sufficient conditions for achieving stability are provided, together with some desired performance in terms of disturbance attenuation. These conditions are expressed in terms of linear programming and scale well with the system size. We then discuss the interconnection of positive subsystems by means of a static output feedback that preserves positivity, and propose conditions to achieve both stability and the asymptotic alignment of the closed-loop output to a desired vector. Finally, we describe some results for a class of parameterized positive systems. The second part of the article presents some interesting applications of the results presented in the first part. Specifically, control problems for heating networks, formation control, power control in wireless communication, and the evolutionary dynamics of cancer and HIV are formalized and solved as optimal control problems for positive systems.

This article explains some fundamental ideas concerning the optimal control of quantum systems through the study of a relatively simple two-level system coupled to optical fields. The model for this system includes both continuous and impulsive dynamics. Topics covered include open- and closed-loop control, impulsive control, open-loop optimal control, quantum filtering, and measurement feedback optimal control.

Reachability analysis consists in computing the set of states that are reachable by a dynamical system from all initial states and for all admissible inputs and parameters. It is a fundamental problem motivated by many applications in formal verification, controller synthesis, and estimation, to name only a few. This article focuses on a class of methods for computing a guaranteed overapproximation of the reachable set of continuous and hybrid systems, relying predominantly on set propagation; starting from the set of initial states, these techniques iteratively propagate a sequence of sets according to the system dynamics. After a review of set representation and computation, the article presents the state of the art of set propagation techniques for reachability analysis of linear, nonlinear, and hybrid systems. It ends with a discussion of successful applications of reachability analysis to real-world problems.

The air transportation system connects the world through the transport of goods and people. However, operational inefficiencies such as flight delays and cancellations are prevalent, resulting in economic and environmental impacts. In the first part of this article, we review recent advances in using network analysis techniques to model the interdependencies observed in the air transportation system and to understand the role of airports in connecting populations, serving air traffic demand, and spreading delays. In the second part, we present some of our recent work on using operational data to build dynamical system models of air traffic delay networks. We show that Markov jump linear system models capture many of the salient characteristics of these networked systems. We illustrate how these models can be validated and then used to analyze system properties such as stability and to design optimal control strategies that limit the propagation of disruptions in air traffic networks.

Network systems consist of subsystems and their interconnections and provide a powerful framework for the analysis, modeling, and control of complex systems. However, subsystems may have high-dimensional dynamics and a large number of complex interconnections, and it is therefore relevant to study reduction methods for network systems. Here, we provide an overview of reduction methods for both the topological (interconnection) structure of a network and the dynamics of the nodes while preserving structural properties of the network. We first review topological complexity reduction methods based on graph clustering and aggregation, producing a reduced-order network model. Next, we consider reduction of the nodal dynamics using extensions of classical methods while preserving the stability and synchronization properties. Finally, we present a structure-preserving generalized balancing method for simultaneously simplifying the topological structure and the order of the nodal dynamics.

We review classic results and recent progress on equilibrium analysis, dynamics, and optimal interventions in network games with both continuous and discrete strategy sets. We study strategic interactions in deterministic networks as well as networks generated from a stochastic network formation model. For the former case, we review a unifying framework for analysis based on the theory of variational inequalities. For the latter case, we highlight how knowledge of the stochastic network formation model can be used by a central planner to design interventions for large networks in a computationally efficient manner when exact network data are not available.

Biomimetic robots that replace living social interaction partners can help elucidate the underlying interaction rules in animal groups. Our review focuses on the use of interactive robots that respond dynamically to animal behavior as part of a closed control loop. We discuss the most influential works to date and how they have contributed to our understanding of animal sociality. Technological advances permit the use of robots that can adapt to the situations they face and the conspecifics they encounter, or robots that learn to optimize their social performance from a set of experiences. We discuss how adaptation and learning may provide novel insights into group sociobiology and describe the technical challenges associatedwith these types of interactive robots. This interdisciplinary field provides a rich set of problems to be tackled by roboticists, machine learning engineers, and control theorists. By cultivating smarter robots, we can usher in an era of more nuanced exploration of animal behavior.

Magnetic microrobotics has undergone approximately 20 years of development, and the robotics and control communities have contributed significant theoretical and practical results to the motion control aspects of this field. This article introduces fundamental motion principles covering individual, multiagent, and swarm control and critically reviews the state of the art along with representative results. It then describes closed-loop control (an important part of this field), including the system structure, current motion planning and control methods, and current feedback approaches. As the development of motion control in magnetic microrobotics is far from complete, especially for swarm control, its current limitations are discussed. Finally, we conclude with several challenges and future research directions.