Out-of-distribution (OOD) detection is the task of identifying inputs that deviate from the training data distribution. This capability is essential for safely deploying deep computer vision models in open-world environments. In this work, we propose a post-hoc method, Perturbation- Rectified OOD detection (PRO), based on the insight that prediction confidence for OOD inputs is more susceptible to reduction under perturbation than in-distribution (IND) inputs. Based on the observation, we propose an adversarial score function that searches for the local minimum scores near the original inputs by applying gradient descent. This procedure enhances the separability between IND and OOD samples. Importantly, the approach improves OOD detection performance without complex modifications to the underlying model architectures. We conduct extensive experiments using the OpenOOD benchmark [43]. Our approach further pushes the limit of softmax-based OOD detection and is the leading post-hoc method for small-scale models. On a CIFAR-10 model with adversarial training, PRO effectively detects near-OOD inputs, achieving a reduction of more than 10% on FPR@95 compared to state-of-the-art methods.1

A correction is given to a previously published result concerned with the relationship between a suitably defined matrix seminorm for consensus analysis and a coefficient of ergodicity.

This paper proposes a novel mission planning algorithm for autonomous robots that selects an optimal waypoint sequence from a predefined set to maximize total reward while satisfying obstacle avoidance, state, input, mission time, and distance constraints. The formulation extends the prize-collecting traveling salesman problem. A tailored genetic algorithm evolves candidate solutions using a fitness function, crossover, and mutation, with constraint enforcement via a penalty method. Differential flatness and clothoid curves are employed to penalize infeasible trajectories efficiently, while the Euler spiral method ensures curvature-continuous trajectories with bounded curvature, enhancing dynamic feasibility and mitigating oscillations typical of minimum-jerk and snap parameterizations. Due to the discrete variable length optimization space, crossover is performed using a dynamic time-warping-based method and extended convex combination with projection. The algorithm’s performance is validated through experiments with a ground vehicle, quadrotor, and quadruped, supported by benchmarking and time-complexity analysis.

We consider the challenging problem of unsupervised change point detection in multivariate time series when the number of change points is unknown. Our method eliminates the user’s need for careful parameter tuning, enhancing its practicality and usability. Our approach identifies time series segments with similar empirically estimated distributions, coupled with a novel greedy algorithm guided by the minimum description length principle. We provide theoretical guarantees and, through experiments on synthetic and real-world data, provide empirical evidence for its improved performance in identifying meaningful change points in practical settings.

This article investigates the problem of communication-efficient and resilient multiagent reinforcement learning (MARL). Specifically, we consider a setting where a set of agents are interconnected over a given network, and can only exchange information with their neighbors. Each agent observes a common Markov Decision Process and has a local cost which is a function of the current system state and the applied control action. The goal of MARL is for all agents to learn a policy that optimizes the infinite horizon discounted average of all their costs. Within this general setting, we consider two extensions to existing MARL algorithms. First, we provide an event-triggered learning rule where agents only exchange information with their neighbors if a certain triggering condition is satisfied. We show that this enables learning while reducing the amount of communication. Next, we consider the scenario where some of the agents can be adversarial (as captured by the Byzantine attack model), and arbitrarily deviate from the prescribed learning algorithm. We establish a fundamental trade-off between optimality and resilience when Byzantine agents are present. We then create a resilient algorithm and show almost sure convergence of all reliable agents' value functions to the neighborhood of the optimal value function of all reliable agents, under certain conditions on the network topology. When the optimal Q -values are sufficiently separated for different actions, we show that all reliable agents can learn the optimal policy under our algorithm.

In this article, we develop the method of continuous Pontryagin differentiable programming (Continuous PDP), which enables a robot to learn an objective function from a few sparsely demonstrated keyframes. The keyframes, labeled with some time stamps, are the desired task-space outputs, which a robot is expected to follow sequentially. The time stamps of the keyframes can be different from the time of the robot's actual execution. The method jointly finds an objective function and a time-warping function such that the robot's resulting trajectory sequentially follows the keyframes with minimal discrepancy loss. The Continuous PDP minimizes the discrepancy loss using projected gradient descent by efficiently solving the gradient of the robot trajectory with respect to the unknown parameters. The method is first evaluated on a simulated robot arm and then applied to a 6-DoF quadrotor to learn an objective function for motion planning in unmodeled environments. The results show the efficiency of the method, its ability to handle time misalignment between keyframes and robot execution, and the generalization of objective learning into unseen motion conditions.

This article proposes a novel approach that enables a robot to learn an objective function incrementally from human directional corrections. Existing methods learn from human magnitude corrections; since a human needs to carefully choose the magnitude of each correction, those methods can easily lead to overcorrections and learning inefficiency. The proposed method only requires human directional corrections—corrections that only indicate the direction of an input change without indicating its magnitude. We only assume that each correction, regardless of its magnitude, points in a direction that improves the robot's current motion relative to an unknown objective function. The allowable corrections satisfying this assumption account for half of the input space, as opposed to the magnitude corrections that have to lie in a shrinking level set. For each directional correction, the proposed method updates the estimate of the objective function based on a cutting plane method, which has a geometric interpretation. We have established theoretical results to show the convergence of the learning process. The proposed method has been tested in numerical examples, a user study on two human–robot games, and a real-world quadrotor experiment. The results confirm the convergence of the proposed method and further show that the method is significantly more effective (higher success rate), efficient/effortless (less human corrections needed), and potentially more accessible (fewer early wasted trials) than the state-of-the-art robot learning frameworks.

In this paper, we develop a distributed control method for an urban traffic network in order to improve traffic conditions and guarantee a smooth operation for all roads and intersections. An optimal coordination among traffic flows is determined by a model predictive control idea. The control objective is formulated as a constrained optimization problem in which the cost function to be minimized is a combination of some traffic network-related performance indexes and the constraints are derived from capacities of roads and intersections. The green time assigned to each road link is computed from its optimal downstream traffic flow. In the proposed algorithms, every road link uses only its local information to determine its control decision which corresponds to a global optimal solution. The algorithms are developed by using a gradient projection-based method and properties of the minimal polynomial of a matrix pair. The effectiveness of the proposed algorithms is validated via numerical simulations with MATLAB and VISSIM.

We propose a Safe Pontryagin Differentiable Programming (Safe PDP) methodology, which establishes a theoretical and algorithmic framework to solve a broad class of safety-critical learning and control tasks---problems that require the guarantee of safety constraint satisfaction at any stage of the learning and control progress. In the spirit of interior-point methods, Safe PDP handles different types of system constraints on states and inputs by incorporating them into the cost or loss through barrier functions. We prove three fundamentals of the proposed Safe PDP: first, both the solution and its gradient in the backward pass can be approximated by solving their more efficient unconstrained counterparts; second, the approximation for both the solution and its gradient can be controlled for arbitrary accuracy by a barrier parameter; and third, importantly, all intermediate results throughout the approximation and optimization strictly respect the constraints, thus guaranteeing safety throughout the entire learning and control process. We demonstrate the capabilities of Safe PDP in solving various safety-critical tasks, including safe policy optimization, safe motion planning, and learning MPCs from demonstrations, on different challenging systems such as 6-DoF maneuvering quadrotor and 6-DoF rocket powered landing.

We consider the problem of modeling piece-wise stationary time series. We propose a new, data-driven technique to automatically identify change-points and learn piece-wise stationary models. We do not assume prior knowledge of the stationary models or the number of change points. Our method can automatically identify repeated stationary models. Our method employs sliding windows and clustering in a novel way. We use the minimum description length principle and integer linear programming to identify the lowest overall complexity system model. Our method does not require parameter tuning and leads to good segmentation and compression. We demonstrate the effectiveness of our method against traditional techniques using both simulated and real-world data.

This paper develops a Pontryagin differentiable programming (PDP) methodology, which establishes a unified framework to solve a broad class of learning and control tasks. The PDP distinguishes from existing methods by two novel techniques: first, we differentiate through Pontryagin's Maximum Principle, and this allows to obtain the analytical derivative of a trajectory with respect to tunable parameters within an optimal control system, enabling end-to-end learning of dynamics, policies, or/and control objective functions; and second, we propose an auxiliary control system in the backward pass of the PDP framework, and the output of this auxiliary control system is the analytical derivative of the original system's trajectory with respect to the parameters, which can be iteratively solved using standard control tools. We investigate three learning modes of the PDP: inverse reinforcement learning, system identification, and control/planning. We demonstrate the capability of the PDP in each learning mode on different high-dimensional systems, including multilink robot arm, 6-DoF maneuvering UAV, and 6-DoF rocket powered landing.

The problem of adaptive fuzzy control is investigated for a class of nontriangular structural stochastic switched nonlinear systems with full state constraints in this paper. A remarkable feature of the nontriangular structural nonlinear system is the so-called algebraic loop problem in the existing backstepping-based analysis and design. Properties of fuzzy basis functions are utilized to circumvent this algebraic loop problem. Based on the Barrier Lyapunov function, an adaptive fuzzy stochastic switched control scheme is designed. It is proven that all the signals in the closed-loop system are semiglobally uniformly ultimately bounded with full state constraints. The effectiveness of the proposed control scheme is verified via simulation studies.

In this paper, we consider a dynamical system whose trajectory is a result of minimizing a multiphase cost function. The multiphase cost function is assumed to be a weighted sum of specified features (or basis functions) with phase-dependent weights that switch at some unknown phase transition points. A new inverse optimal control approach for recovering the cost weights of each phase and estimating the phase transition points is proposed. The key idea is to use a length-adapted window moving along the observed trajectory, where the window length is determined by finding the minimal observation length that suffices for a successful cost weight recovery. The effectiveness of the proposed method is first evaluated on a simulated robot arm, and then, demonstrated on a dataset of human participants performing a series of squatting tasks. The results demonstrate that the proposed method reliably retrieves the cost function of each phase and segments each phase of motion from the trajectory with a segmentation accuracy above 90%.

In this paper, an adaptive fault estimation approach is proposed for a class of switched nonlinear systems. The considered system is assumed to possess unknown nonlinearities, unmeasured states, and simultaneous sensor and actuator faults. Fuzzy logic systems are applied to approximate the unknown nonlinear terms. Two new adaptive fuzzy observers are designed where the sensor and actuator faults can be estimated separately. On the basis of average dwell time approach and Lyapunov stability theory, the resulting error system is proven to be bounded stable with designed parameters. Finally, a simulation example is presented to illustrate the effectiveness of the designed observers.

In Fauré et al. (2006), the dynamics of the core network regulating the mammalian cell cycle is formulated as a Boolean control network (BCN) model consisting of nine proteins as state nodes and a tenth protein (protein CycD) as the control input node. In this model, one of the state nodes, protein Cdc20, plays a central role in the separation of sister chromatids. Hence, if any Cdc20 sequence can be obtained, fully controlling the mammalian cell cycle is feasible. Motivated by this fact, we study whether any Cdc20 sequence can be obtained theoretically. We formulate the foregoing problem as the invertibility of BCNs, that is, whether one can obtain any Cdc20 sequence by designing input (i.e., protein CycD) sequences. We give an algorithm to verify the invertibility of any BCN, and find that the BCN model for the core network regulating the mammalian cell cycle is not invertible, that is, one cannot obtain any Cdc20 sequence. We further present another algorithm to test whether a finite Cdc20 sequence can be generated by the BCN model, which leads to a series of periodic infinite Cdc20 sequences with alternately active and inactive Cdc20 segments. States of these sequences are alternated between the two attractors in the proposed model, which reproduces correctly how a cell exits the cell cycle to enter the quiescent state, or the opposite.

Some applications of the interconnected embedded systems such as sensor networks rely on all nodes in the network to execute certain tasks simultaneously. To meet this demand for simultaneity, a multi-timer model based fully distributed task synchronization algorithm is proposed in this paper. In this multi-timer model, each node containing an embedded system is characterized by a timer. The microcontrollers (MCUs) within the interconnected embedded systems are switched to the assigned tasks by timer interrupts. Each timer decides when to trigger interrupts by only using the information from its neighbors. Task synchronization is realized by using the proposed synchronization algorithm. Some simulation examples are presented in the end to verify the effectiveness of the proposed synchronization algorithm.

For the purposes of this paper, “gossiping” is a distributed process whose purpose is to enable the members of a group of autonomous agents to asymptotically determine, in a decentralized manner, the average of the initial values of their scalar gossip variables. This paper discusses several different deterministic protocols for gossiping which avoid deadlocks and achieve consensus under different assumptions. First considered is$T$-periodic gossiping which is a gossiping protocol which stipulates that each agent must gossip with the same neighbor exactly once every$T$time units. Among the results discussed is the fact that if the underlying graph characterizing neighbor relations is a tree, convergence is exponential at a worst case rate which is the same for all possible$T$-periodic gossip sequences associated with the graph. Many gossiping protocols are request based which means simply that a gossip between two agents will occur whenever one of the two agents accepts a request to gossip placed by the other. Three deterministic request-based protocols are discussed. Each is guaranteed to not deadlock and to always generate sequences of gossip vectors which converge exponentially fast. It is shown that worst case convergence rates can be characterized in terms of the second largest singular values of suitably defined doubly stochastic matrices.

In this brief, the problem of global asymptotic stability for delayed Hopfield neural networks (HNNs) is investigated. A new criterion of asymptotic stability is derived by introducing a new kind of Lyapunov-Krasovskii functional and is formulated in terms of a linear matrix inequality (LMI), which can be readily solved via standard software. This new criterion based on a delay fractioning approach proves to be much less conservative and the conservatism could be notably reduced by thinning the delay fractioning. An example is provided to show the effectiveness and the advantage of the proposed result.

In this correspondence, the problem of exponential stability for neural networks with time delay is investigated. By introducing a novel Lyapunov-Krasovskii functional with the idea of delay fractioning, a new criterion of exponential stability is derived and then formulated in terms of a linear matrix inequality. This new criterion proves to be much less conservative than the most recent result, and the conservatism can be notably reduced as the fractioning goes thinner. An example is provided to demonstrate the advantage of the proposed result.