Determining which dynamical systems exhibit behavior most similar to a target system is critical for transfer learning and data-efficient system identification. However, quantifying similarity between nonlinear systems from input-output data is challenging: trajectories differ in length, initial conditions, and input sequences, making direct comparison unreliable. Classical system identification depends on correct model structure selection, while frequency-domain methods often miss transient dynamics. This paper proposes a data-driven methodology based on matrix profiles: a time-series tool for efficient subsequence matching. The approach operates directly on input-output trajectories without state estimation or parameter identification. By embedding trajectories in an interleaved input-output space and computing cross-matrix profiles, the method captures dynamical features including transient behavior under varying conditions. We derive analytical bounds on matrix profile distance as functions of parameter and input discrepancies, establishing that systems closer in parameter space yield smaller matrix profile distances when control trajectories are similar. Numerical experiments demonstrate improved similarity assessment compared to Euclidean distance and dynamic time warping baselines, achieving substantial reductions in trajectory error and parameter-space deviation across both linear and nonlinear systems.

The paper presents a method for learning state representations of articulated mechanisms consisting of multiple links connected by joints but not equipped with positional encoders, from short sequences of visual observations of the mechanism collected by stationary RGB-D cameras. The method leverages recent advances in algorithms based on deep neural networks for long-term keypoint tracking in image sequences to extract tracks of matching keypoints over time, even when some of them are temporarily occluded. Because keypoint tracking is not perfectly reliable, we propose a novel method for robust clustering to discover how many rigid bodies (links) the mechanism consists of by analyzing which points move together and thus must belong to the same rigid body. Experiments with recently proposed long-term keypoint trackers demonstrate successful state estimation of the kinematic structure of articulated mechanisms entirely from camera images, effectively allowing computation of inverse kinematics and visual servocontrol for mechanisms such as robot arms, cranes, etc. even without positional encoders.

The paper proposes a method for visual servo control of nonholonomic robots with unknown dynamics using images captured by uncalibrated cameras. The method learns the transition dynamics of the robot directly in visual feature space and linearizes it successively in order to compute controls. Experiments both in simulation and on a real testbed using a unicycle-type mobile robot demonstrate that the use of planning and trajectory stabilization algorithms based on differential dynamic programming is much more effective in handling nonholonomic constraints in executing difficult maneuvers, such as parallel parking, than more traditional visual servoing schemes that linearize the learned dynamics around a single point. The ability of the proposed method to control nonholonomic robots without manually calibrating cameras and identifying robot dynamics could potentially significantly lower the cost of deployment of autonomous mobile robots at scale.

The paper proposes a method for learning compact representations of the configuration (joint positions) of articulated mechanisms consisting of interconnected rigid bodies, from collected sequences of keypoint positions observed and tracked in camera images. The method analyzes the variations in pairwise distances between keypoints over time to deduce which of the keypoints must belong to the same rigid body and then computes the relative pose of all rigid bodies with respect to a reference image representing an initial or target configuration of the mechanism. By analyzing the rank of data matrices representing the translational and rotational components of the relative poses between the rigid bodies over time, the algorithm infers the order of the kinematic chain of the mechanism and the type of joints used in it, allowing the construction of a configuration vector as compact as the true joint positions of the mechanism.

We propose a method for visual servo-control of robots using images from an uncalibrated camera that constructs compact state representations of the robot’s configuration and uses transition dynamics learned from collected execution traces to compute control velocities to reach a desired goal state identified directly by its image. The key step of the proposed method is the estimation of a homography transform between the image positions of distinct keypoints belonging to the robot in the current image and those in a reference image, which can be done quickly and robustly even when not the same set of keypoints is observed at each time step, making it robust to noise and variations in illumination. The estimated homography is then used to represent the robot configuration as the image coordinates of a minimal number of virtual points moving with the robot. The method was verified experimentally for planar motion of a fully actuated manipulator arm as well as an underactuated mobile robot with a nonholonomic constraint.

We propose a method for designing global nonlinear controllers based on the application of memory-based learning schemes for the purpose of aggregating multiple solutions produced by optimal control algorithms based on differential dynamic programming. The method leverages the fact that these optimal control algorithms produce not only nominal state and control trajectories, but entire full-state feedback (FSF) controllers, and the combined controller effectively switches between these multiple FSF controllers. Empirical verification demonstrates that it can be very effective in solving difficult benchmark control problems at high control rates.

The paper proposes a method for learning velocity estimators, in the form of finite impulse response (FIR) filters, from data collected from a system equipped with quantizing position encoders that is to be controlled by means of a full-state feedback controller making use of the velocity estimates. The resulting estimators are tailored to the properties of the controlled system and show empirically superior performance in comparison with commonly used baseline velocity estimators, both in terms of velocity estimation error as well as in terms of reduced regulation cost when tested on control problems. The proposed adaptive estimators are resistant to overfitting the training data, are easy to implement on embedded controller devices, and can be used in conjunction with various learning control methods.

Extreme weather events have posed tremendous challenges to the operation of distribution networks. In this paper, we propose a decision-dependent chance-constrained model for the optimal planning of diesel generators, renewable distributed generations (RDGs), energy storage systems, and switches under contingency. A promising moment-based ambiguity set that incorporates the information of decision variables is employed to depict the uncertainty arising from RDGs. By leveraging effective approximation methods such as the Bonferroni approximation method to handle the considered joint chance constraints, the proposed model is transformed into a tractable mixed-integer second-order conic programming problem, which means it can easily be implemented. Numerical experiments are put forward on the IEEE 33-bus test system to validate the effectiveness of the developed approach.

We present a method for designing and tuning controllers for the problem of swing-up and stabilization of a Furuta pendulum. The method is based on suitable parameterization of a family of controllers and the application of Bayesian optimization to their tuning with minimal interaction with the physical system. Unlike traditional controller design methodologies, the method does not require the derivation of an exact physical model of the controlled plant, thus saving significant design time and effort. Furthermore, the method has much more favorable sample complexity than most policy optimization methods proposed in the field of reinforcement learning.

For more than half a century, vibratory bowl feeders have been the standard in automated assembly for singulation, orientation, and manipulation of small parts. Unfortunately, these feeders are expensive, noisy, and highly specialized on a single part design bases. We consider an alternative device and learning control method for singulation, orientation, and manipulation by means of seven fixed-position variable-energy solenoid impulse actuators located beneath a semi-rigid part supporting surface. Using computer vision to provide part pose information, we tested various machine learning (ML) algorithms to generate a control policy that selects the optimal actuator and actuation energy. Our manipulation test object is a 6-sided craps-style die. Using the most suitable ML algorithm, we were able to flip the die to any desired face 30.4% of the time with a single impulse, and 51.3% with two chosen impulses, versus a random policy succeeding 5.1% of the time (that is, a randomly chosen impulse delivered by a randomly chosen solenoid).

Robots have been steadily increasing their presence in our daily lives, where they can work along with humans to provide assistance in various tasks on industry floors, in offices, and in homes. Automated assembly is one of the key applications of robots, and the next generation assembly systems could become much more efficient by creating collaborative human-robot systems. However, although collaborative robots have been around for decades, their application in truly collaborative systems has been limited. This is because a truly collaborative human-robot system needs to adjust its operation with respect to the uncertainty and imprecision in human actions, ensure safety during interaction, etc. In this paper, we present a system for human-robot collaborative assembly using learning from demonstration and pose estimation, so that the robot can adapt to the uncertainty caused by the operation of humans. Learning from demonstration is used to generate motion trajectories for the robot based on the pose estimate of different goal locations from a deep learning-based vision system. The proposed system is demonstrated using a physical 6 DoF manipulator in a collaborative human-robot assembly scenario. We show successful generalization of the system's operation to changes in the initial and final goal locations through various experiments.

We propose a combined particle-based density prediction model consisting of three components: trajectory prediction for existing particles, entering particle prediction, and iterative sampling. At initialization, the combined model takes in a set of trajectories for trajectory prediction and a sequence of observation vectors for entering particle prediction. Then, the iterative sampling module generates the density prediction for the next time instance. It will also sample a pool of particles and pass on their trajectories to the next trajectory prediction model for future density prediction.

In this paper, we propose to estimate the forward dynamics equations of mechanical systems by learning a model of the inverse dynamics and estimating individual dynamics components from it. We revisit the classical formulation of rigid body dynamics in order to extrapolate the physical dynamical components, such as inertial and gravitational components, from an inverse dynamics model. After estimating the dynamical components, the forward dynamics can be computed in closed form as a function of the learned inverse dynamics. We tested the proposed method with several machine learning models based on Gaussian Process Regression and compared them with the standard approach of learning the forward dynamics directly. Results on two simulated robotic manipulators, a PANDA Franka Emika and a UR10, show the effectiveness of the proposed method in learning the forward dynamics, both in terms of accuracy as well as in opening the possibility of using more structured models.

Insertion operations are a critical element of most robotic assembly operation, and peg-in-hole (PiH) insertion is one of the most widely studied tasks in the industrial and academic manipulation communities. PiH insertion is in fact an entire class of problems, where the complexity of the problem can depend on the type of misalignment and contact formation during an insertion attempt. In this paper, we present the design and analysis of adaptive compliance controllers which can be used in insertion-type assembly tasks, including learning-based compliance controllers which can be used for insertion problems in the presence of uncertainty in the goal location during robotic assembly. We first present the design of compliance controllers which can ensure safe operation of the robot by limiting experienced contact forces during contact formation. Consequently, we present analysis of the force signature obtained during the contact formation to learn the corrective action needed to perform insertion. Finally, we use the proposed compliance controllers and learned models to design a policy that can successfully perform insertion in novel test conditions with almost perfect success rate. We validate the proposed approach on a physical robotic test-bed using a 6-DoF manipulator arm.

Accurate estimation of travel times is an important step in smart transportation and smart building systems. Poor estimation of travel times results in both frustrated users and wasted resources. Current methods that estimate travel times usually only return point estimates, losing important distributional information necessary for accurate decision-making. We propose using neural network-based mixture distributions to predict a user's travel times given their origin and destination coordinates. We show that our method correctly estimates the travel time distribution, maximizes utility in a downstream elevator scheduling task, and is easy to retrain—making it a versatile and an inexpensive-to-maintain module when deployed in smart crowd management systems.

The need for accurate and timely destination prediction arises in many transportation applications. We formulate destination prediction as a multivariate time series classification problem, and leverage part of the core components of the Transformer network to build a new deep neural network model exclusively for this task. The key building block of our model consists of Two Towers of Transformer encoders, and we call it “3T-Net.” Through extensive comparison experiments on a simulated indoor trajectories data set, we show that 3T-Net performs better or close to other investigated state-of-the-art deep learning based models. Our model can also be used for outdoor destination prediction scenarios and more general multivariate time series classification problems.

Detecting and locating High Impedance Faults (HiZ) is difficult due to the small magnitude of fault current during such faults. In this work, we propose a combined HiZ fault detection and localization technique that uses voltage and current measurements available from existing intelligent electronic devices (IEDs). At first, we apply the variation mode decomposition (VMD) model to detect the existence of fault based on denoised time series of measurements using Wavelet Transform (WT). After detecting the presence of fault, we apply the correlation based matrix to locate the suspicious fault locations, and then utilize K-nearest neighbour (KNN) to identify the faulty branch among those locations using dynamic time warping method to measure the distance between neighbors. Finally, we verify our proposed model with simulation results on an inverter-based microgrid. Outcomes from VMD method demonstrate that measurement from any location of the grid can indicate the existence of HiZ fault over the duration of the event, and show the scalability of the proposed method. For localization, it is verified that the correlation matrix combined with KNN can remove the false positive cases with properly tuned KNN-parameters and correlation matrix threshold, irrespective of the measurement noise.

Dynamic movement primitives are widely used for learning skills that can be demonstrated to a robot by a skilled human or controller. While their generalization capabilities and simple formulation make them very appealing to use, they possess no strong guarantees to satisfy operational safety constraints for a task. We present constrained dynamic movement primitives (CDMPs), which can allow for positional constraint satisfaction in the robot workspace. Our method solves a non-linear optimization to perturb an existing DMP's forcing weights to admit a Zeroing Barrier Function (ZBF), which certifies positional workspace constraint satisfaction. We demonstrate our approach under different positional constraints on the end-effector movement on multiple physical robots, such as obstacle avoidance and workspace limitations.