Stand-alone Visual Place Recognition (VPR) systems have little defence against a well-designed adversarial attack, which can lead to disastrous consequences when deployed for robot navigation. This paper extensively analyzes the effect of two adversarial attacks common in other perception tasks and two novel VPR-specific attacks on VPR localization performance. We then propose how to close the loop between VPR, an Adversarial Attack Detector (AAD), and active navigation decisions by demonstrating the performance benefit of simulated AADs in a novel experiment paradigm – which we detail for the robotics community to use as a system framework. In the proposed experiment paradigm, we see the addition of AADs across a range of detection accuracies can improve performance over baseline; demonstrating a significant improvement – such as a ≈ 50% reduction in the mean along-track localization error – can be achieved with True Positive and False Positive detection rates of only 75% and up to 25% respectively. We examine a variety of metrics including: Along-Track Error, Percentage of Time Attacked, Percentage of Time in an ‘Unsafe’ State, and Longest Continuous Time Under Attack. Expanding further on these results, we provide the first investigation into the efficacy of the Fast Gradient Sign Method (FGSM) adversarial attack for VPR. The analysis in this work highlights the need for AADs in real-world systems for trustworthy navigation, and informs quantitative requirements for system design.

In this paper we characterise the long-run behaviour of the replicator dynamic in zero-sum games (symmetric or non-symmetric). Specifically, we prove that every zero-sum game possesses a unique global replicator attractor, which we then characterise. Most surprisingly, this attractor depends only on each player’s preference order over their own strategies and not on the cardinal payoff values, defined by a finite directed graph we call the game’s preference graph. When the game is symmetric, this graph is a tournament whose nodes are strategies; when the game is not symmetric, this graph is the game’s response graph. We discuss the consequences of our results on chain recurrence and Nash equilibria.

In this paper we examine the relationship between the flow of the replicator dynamic, the continuum limit of Multiplicative Weights Update, and a game’s \emph{response graph}. We settle an open problem establishing that under the replicator, \emph{sink chain components}—a topological notion of long-run outcome of a dynamical system—always exist and are approximated by the \emph{sink connected components} of the game’s response graph. More specifically, each sink chain component contains a sink connected component of the response graph, as well as all mixed strategy profiles whose support consists of pure profiles in the same connected component, a set we call the \emph{content} of the connected component. As a corollary, all profiles are chain recurrent in games with strongly connected response graphs. In any two-player game sharing a response graph with a zero-sum game, the sink chain component is unique. In two-player zero-sum and potential games the sink chain components and sink connected components are in a one-to-one correspondence, and we conjecture that this holds in all games.

Robust autonomy stacks require tight integration of perception, motion planning, and control layers, but these layers often inadequately incorporate inherent perception and prediction uncertainties, either ignoring them altogether or making questionable assumptions of Gaussianity. Robots with nonlinear dynamics and complex sensing modalities operating in an uncertain environment demand more careful consideration of how uncertainties propagate across stack layers. We propose a framework to integrate perception, motion planning, and control by explicitly incorporating perception and prediction uncertainties into planning so that risks of constraint violation can be mitigated. Specifically, we use a nonlinear model predictive control based steering law coupled with a decorrelation scheme based Unscented Kalman Filter for state and environment estimation to propagate the robot state and environment uncertainties. Subsequently, we use distributionally robust risk constraints to limit the risk in the presence of these uncertainties. Finally, we present a layered autonomy stack consisting of a nonlinear steering-based distributionally robust motion planning module and a reference trajectory tracking module. Our numerical experiments with nonlinear robot models and an urban driving simulator show the effectiveness of our proposed approaches.

With advances in image processing and machine learning, it is now feasible to incorporate semantic information into the problem of simultaneous localisation and mapping (SLAM). Previously, SLAM was carried out using lower level geometric features (points, lines, and planes) which are often view-point dependent and error prone in visually repetitive environments. Semantic information can improve the ability to recognise previously visited locations, as well as maintain sparser maps for long term SLAM applications. However, SLAM in repetitive environments has the critical problem of assigning measurements to the landmarks which generated them. In this paper, we use k-best assignment enumeration to compute marginal assignment probabilities for each measurement landmark pair, in real time. We present numerical studies on the KITTI dataset to demonstrate the effectiveness and speed of the proposed framework.

In this letter, we consider the problem of field estimation using binary measurements. Previous work has formulated the problem as a parameter estimation problem, which can be solved in an online manner using sequential Monte Carlo (SMC) techniques. In the current work, we consider an alternative approach to the parameter estimation problem based on online logistic regression. The developed algorithm is less computationally intensive than the SMC approach, while having more reliable estimation performance.

Modularity is a central principle throughout the design process for cyber-physical systems. Modularity reduces complexity and increases reuse of behavior. In this article we pose and answer the following question: how can we identify independent “modules” within the structure of reactive control architectures? To this end, we propose a graph-structured control architecture we call a decision structure and show how it generalizes some reactive control architectures that are popular in Artificial Intelligence (AI) and robotics, specifically Teleo-Reactive programs (TRs), Decision Trees (DTs), Behavior Trees (BTs), and Generalised Behavior Trees ( k -BTs). Inspired by the definition of a module in graph theory [ 16 ] we define modules in decision structures and show how each decision structure possesses a canonical decomposition into its modules, which can be found in polynomial time. We establish intuitive connections between our proposed modularity and modularity in structured programming. In BTs, k -BTs, and DTs the modules we propose are in a one-to-one correspondence with their subtrees. We show we can naturally characterize each of the BTs, k -BTs, DTs, and TRs by properties of their module decomposition. This allows us to recognize which decision structures are equivalent to each of these architectures in quadratic time. Following McCabe [ 26 ], we define a complexity measure called essential complexity on decision structures, which measures the degree to which they can be decomposed into simpler modules. We characterize the k -BTs as the decision structures of unit-essential complexity. Our proposed concept of modules extends to formal verification, under any verification scheme capable of verifying a decision structure. Namely, we prove that a modification to a module within a decision structure has no greater flow-on effects than a modification to an individual action within that structure. This enables verification on modules to be done locally and hierarchically, where structures can be verified and then repeatedly locally modified, with modules replaced by modules while preserving correctness. To illustrate the findings, we present an example of a solar-powered drone completing a reconnaissance-based mission using a decision structure. We use a Linear Temporal Logic-based verification scheme to verify the correctness of this structure and then show how one can repeatedly modify modules while preserving its correctness, and this can be verified by considering only those modules that have been modified.

We address the problem of maximizing privacy of stochastic dynamical systems whose state information is released through quantized sensor data. In particular, we consider the setting where information about the system state is obtained using noisy sensor measurements. This data is quantized and transmitted to a (possibly untrustworthy) remote station through a public/unsecured communication network. We aim at keeping (part of) the state of the system private; however, because the network (and/or the remote station) might be unsecure, adversaries might have access to sensor data, which can be used to estimate the system state. To prevent such adversaries from obtaining an accurate state estimate, before transmission, we randomize quantized sensor data using additive random vectors, and send the corrupted data to the remote station instead. We design the joint probability distribution of these additive vectors (over a time window) to minimize the mutual information (our privacy metric) between some linear function of the system state (a desired private output) and the randomized sensor data for a desired level of distortion-how different quantized sensor measurements and distorted data are allowed to be. We pose the problem of synthesising the joint probability distribution of the additive vectors as a convex program subject to linear constraints. Simulation experiments are presented to illustrate our privacy scheme.

We propose an algorithm for generating minimum-snap trajectories for quadrotors with linear computational complexity with respect to the number of segments in the spline trajectory. Our algorithm is numerically stable for large numbers of segments and is able to generate trajectories of more than 500, 000 segments. The computational speed and numerical stability of our algorithm makes it suitable for real-time generation of very large scale trajectories. We demonstrate the performance of our algorithm and compare it to existing methods, in which it is both faster and able to calculate larger trajectories than state-of-the-art. We also show the feasibility of the trajectories experimentally with a long quadrotor flight.

Many industrial domains are characterized by Multiple-Input-Multiple-Output (MIMO) systems for which an explicit relationship capturing the nontrivial trade-off between the competing objectives is not available. Human experts have the ability to implicitly learn such a relationship, which in turn enables them to tune the corresponding controller to achieve the desirable closed-loop performance. However, as the complexity of the MIMO system and/or the controller increase, so does the tuning time and the associated tuning cost. To reduce the tuning cost, a framework is proposed in which a machine learning method for approximating the human-learned cost function along with an optimization algorithm for optimizing it, and consequently tuning the controller, are employed. In this work the focus is on the tuning of Model Predictive Controllers (MPCs), given both the interest in their implementations across many industrial domains and the associated high degrees of freedom present in the corresponding tuning process. To demonstrate the proposed approach, simulation results for the tuning of an air path MPC controller in a diesel engine are presented.

In this letter, the interplay between a class of nonlinear estimators and strategic sensors is studied in several participatory-sensing scenarios. It is shown that for the class of estimators, if the strategic sensors have access to noiseless measurements of the to-be-estimated-variable, truth-telling is an equilibrium of the game that models the interplay between the sensors and the estimator. Furthermore, performance of the proposed estimators is examined in the case that the strategic sensors form coalitions and in the presence of noise.

Dual control aims to concurrently learn and control an unknown system. However, actively learning the system conflicts directly with any given control objective for it will disturb the system during exploration. This paper presents a receding horizon approach to dual control, where a multiobjective optimization problem is solved repeatedly and subject to constraints representing system dynamics. Balancing a standard finite-horizon control objective, a knowledge gain objective is defined to explicitly quantify the information acquired when learning the system dynamics. Measures from information theory, such as entropy-based uncertainty, Fisher information, and relative entropy, are studied and used to quantify the knowledge gained as a result of the control actions. The resulting iterative framework is applied to Markov decision processes and discrete-time nonlinear systems. Thus, the broad applicability and usefulness of the presented approach is demonstrated in diverse problem settings. The framework is illustrated with multiple numerical examples.

The ability to maintain state awareness in the face of unexpected and unmodeled errors and threats is a defining feature of a resilient control system. Therefore, in this paper, we study the problem of distributed fault detection and isolation (FDI) in large networked systems with uncertain system models. The linear networked system is composed of interconnected subsystems and may be represented as a graph. The subsystems are represented by nodes, while the edges correspond to the interconnections between subsystems. Considering faults that may occur on the interconnections and subsystems, as our first contribution, we propose a distributed scheme to jointly detect and isolate faults occurring in nodes and edges of the system. As our second contribution, we analyze the behavior of the proposed scheme under model uncertainties caused by the addition or removal of edges. Additionally, we propose a novel distributed FDI scheme based on local models and measurements that is resilient to changes outside of the local subsystem and achieves FDI. Our third contribution addresses the complexity reduction of the distributed FDI method, by characterizing the minimum amount of model information and measurements needed to achieve FDI and by reducing the number of monitoring nodes. The proposed methods can be fused to design a scalable and resilient distributed FDI architecture that achieves local FDI despite unknown changes outside the local subsystem. The proposed approach is illustrated by numerical experiments on the IEEE 118-bus power network benchmark.

This paper addresses a surveillance problem in which the goal is to achieve a circular motion around a target by a non-holonomic agent. The agent only knows its own position with respect to its initial frame, and the bearing angle of the target in that frame. It is assumed that the position of the target is unknown. An estimator and a controller are proposed to estimate the position of the target and make the agent move on a circular trajectory with a desired radius around it. The performance of the proposed algorithm is verified both through simulations and experiments. Robustness is also established in the face of noise and target motion.

In this letter, we study the problem of fault detection and mitigation in networks where the measurements satisfy Kirchhoff's voltage law. First, we characterise the class of faults appearing as an additive fault vector (injected by a malicious adversary or due to equipment failures) that can be detected by taking into account the topology of the network. Second, we consider the problem of estimating the fault vector via tools from compressive sensing. Moreover, we comment on the applicability of the developed methods to the case where the measurements satisfy Kirchhoff's current law. The proposed methods are validated via numerical examples with application to time synchronization networks.

Graph theory has been used to characterize the solvability of the sensor network localization problem. If sensors correspond to vertices and edges correspond to sensor pairs between which the distance is known, a significant result in the theory of range-based sensor network localization is that if the graph underlying the sensor network is generically globally rigid and there is a suitable set of anchors at known positions, then the network can be localized, i.e., a unique set of sensor positions can be determined that is consistent with the data. In particular, for planar problems, provided the sensor network has three or more noncollinear anchors at known points, all sensors are located at generic points, and the intersensor distances corresponding to the graph edges are precisely known rather than being subject to measurement noise, generic global rigidity of the graph is necessary and sufficient for the network to be localizable (in the absence of any further information). In practice, however, distance measurements will never be exact, and the equations whose solutions deliver sensor positions in the noiseless case in general no longer have a solution. This paper then argues that if the distance measurement errors are not too great and otherwise the associated graph is generically globally rigid and there are three or more noncollinear anchors, the network will be approximately localizable, in the sense that estimates can be found for the sensor positions which are near the correct values; in particular, a bound on the position errors can be found in terms of a bound on the distance errors. The sensor positions in this case can be found by minimizing a cost function which, although nonconvex, does have a global minimum.

In this paper, we have taken advantage of reinforcement learning to develop a new traffic shaper in order to obtain a reasonable utilization of bandwidth while preventing traffic overload in other part of the network and as a result, reducing total number of packet dropping in the whole network.. We used a modified version of Q-learning in which a combination of neural networks keeps the data of Q-table in order to make the operation faster while keeping the required storage as small as possible. This method shows satisfactory results in simulations from the aspects of keeping dropping probability low while injecting as many packets as possible into the network in order to utilize the free bandwidth as much as possible. On the other hand the results show that the system can perform in situations that are not originally designed to act in

In this paper, a new intelligent traffic shaper is proposed to obtain a reasonable utilization of bandwidth while preventing traffic overload in other part of the network and as a result, reducing total number of packet dropping in the whole network. This approach trains an intelligent agent to learn an appropriate value for token generation rate of a Token Bucket at various states of the network. This method shows satisfactory results in simulations from the aspects of keeping dropping probability low while injecting as many packets as possible into the network by minimization of used buffer size at each router in order to keep the delay occurred by packets waiting in long buffers to be sent, as small as possible