Hard Constraints Research Articles

Biologically informed NeuralODEs for genome-wide regulatory dynamics

BackgroundGene regulatory network (GRN) models that are formulated as ordinary differential equations (ODEs) can accurately explain temporal gene expression patterns and promise to yield new insights into important cellular processes, disease progression, and intervention design. Learning such gene regulatory ODEs is challenging, since we want to predict the evolution of gene expression in a way that accurately encodes the underlying GRN governing the dynamics and the nonlinear functional relationships between genes. Most widely used ODE estimation methods either impose too many parametric restrictions or are not guided by meaningful biological insights, both of which impede either scalability, explainability, or both.ResultsWe developed PHOENIX, a modeling framework based on neural ordinary differential equations (NeuralODEs) and Hill-Langmuir kinetics, that overcomes limitations of other methods by flexibly incorporating prior domain knowledge and biological constraints to promote sparse, biologically interpretable representations of GRN ODEs. We tested the accuracy of PHOENIX in a series of in silico experiments, benchmarking it against several currently used tools. We demonstrated PHOENIX’s flexibility by modeling regulation of oscillating expression profiles obtained from synchronized yeast cells. We also assessed the scalability of PHOENIX by modeling genome-scale GRNs for breast cancer samples ordered in pseudotime and for B cells treated with Rituximab.ConclusionsPHOENIX uses a combination of user-defined prior knowledge and functional forms from systems biology to encode biological “first principles” as soft constraints on the GRN allowing us to predict subsequent gene expression patterns in a biologically explainable manner.

Constructing topology‐constrained distance cartograms with application on spatial interaction data

AbstractA distance cartogram is a deformed map where the distance between points conforms to a specific proximity indicator. Its readability is crucial, requiring a similar spatial arrangement of points between the original map and cartogram. Previous studies mainly incorporated angle changes of point pairs into the optimization objective. However, this soft constraint fails to provide high readability for spatial interaction data with numerous points and links. This study emphasizes the significance of maintaining Delaunay triangulation during deformation. To achieve this, topology‐constrained particle swarm optimization (TC‐PSO) is proposed, in which triangle intersections and flipping are prevented during optimization. Additionally, a topology error is introduced to evaluate the difference in triangulation between the original and deformed maps. TC‐PSO outperforms previous approaches by exhibiting the smallest topology error and producing more readable cartograms in simulation experiments and Baidu index data. These show TC‐PSO's advantage as a cartographic tool.

Ergodic mean field games: existence of local minimizers up to the Sobolev critical case

We investigate the existence of solutions to viscous ergodic Mean Field Games systems in bounded domains with Neumann boundary conditions and local, possibly aggregative couplings. In particular we exploit the associated variational structure and search for constrained minimizers of a suitable functional. Depending on the growth of the coupling, we detect the existence of global minimizers in the mass subcritical and critical case, and of local minimizers in the mass supercritical case, notably up to the Sobolev critical case.

Leveraging Organizational Hierarchy to Simplify Reward Design in Cooperative Multi-agent Reinforcement Learning

The effectiveness of multi-agent reinforcement learning (MARL) hinges largely on the meticulous arrangement of objectives. Yet, conventional MARL methods might not completely harness the inherent structures present in environmental states and agent relationships for goal organization. This study is conducted within the domain of military training simulations, which are typically characterized by complexity, heterogeneity, non-stationary and doctrine-driven environments with a clear organizational hierarchy and a top-down chain of command. This research investigates the approximation and integration of the organizational hierarchy into MARL for cooperative training scenarios, with the goal of streamlining the processes of reward engineering and enhancing team coordination. In the preliminary experiments, we employed two-tiered commander-subordinate feudal hierarchical (CSFH) networks to separate the prioritized team goal and individual goals. The empirical results demonstrate that the proposed framework enhances learning efficiency. It guarantees the learning of a prioritized policy for the commander agent and encourages subordinate agents to explore areas of interest more frequently, guided by appropriate soft constraints imposed by the commander.

Constrained Tiny Machine Learning for Predicting Gas Concentration with I4.0 Low-cost Sensors

Low-cost gas sensors (LCS) often produce inaccurate measurements due to varying environmental conditions that are not consistent with laboratory settings, leading to inadequate productivity levels compared to high-quality sensors. To address this issue, we propose the use of Machine Learning (ML) to predict accurate concentrations of pollutant gases acquired by LCS integrated into an embedded Internet of Things platform. However, a key challenge is to optimize an accurate ML design under low memory and computation power constraints of microcontrollers (MCUs) while maintaining accurate ML scores. After data analysis and pre-processing, we assess and analyze the performance of five ML algorithms to predict the concentration of pollutants gases from multiple specifications (weather, presence of other gases, etc.). To support the experiments, datasets from three sources are used: (1) VOCSens, (2) Belgian Interregional Environment Agency cell, and (3) Visual-Crossing. Once the best model was optimized and validated, multiple hard constraints were added to the selected ML structure to satisfy material and expert requirements. Trained models were ported to be implemented locally in a MCU after comparing several porting libraries. The assembled code obtained is evaluated based on two metrics: storage memory consumption and inference time, relative to the highest attainable capacities. The improved random forest is the best ML model for the used dataset with an R2 score meeting of 0.72 and Root Means Square Error of 0.0028 ppm. The best generated Tiny-ML model needs 3% of RAM and 98% of Flash storage. The empirical results prove that the developed ML algorithm applied to LCS provides high accuracy to predict pollutant gases. This algorithm can also be used to adjust the LCS systems to provide calibrated data in real time, even if the platform being used is not particularly advanced or powerful.

Geographic coordinate validation and assignment using an edge-constrained layout

Electric grids with buses that are mapped to geographic latitude and longitude are useful for a growing number of applications, such as data visualization, geomagnetically induced current calculations, and multi-energy coupled infrastructure simulations. This paper presents a methodology for validating the quality of geographic coordinates for a power system model, and to assign coordinates to buses with missing or low-quality coordinates. This method takes advantage of geographic indicators already intrinsic to a grid model, such as branch length as implied by impedance and susceptance parameters. The coordinate assignment process uses an approach inspired by graph drawing, that lays out the vertices (buses) and edges (transmission lines), formulated as a nonlinear programming problem with soft edge length constraints. The layout method is very computationally fast and scalable to large power system cases. The method is demonstrated in this paper using a 37-bus test case and a 6717-bus test case, both publicly available, along with a large actual grid model. The results show that, for cases with only a few errors in the coordinates, cases with no coordinates known beforehand, and others in between, this method is able to assign reasonable geographic coordinates to best match known data about the grid.

Influence of learned landmark correspondences on lung CT registration.

Disease or injury may cause a change in the biomechanical properties of the lungs, which can alter lung function. Image registration can be used to measure lung ventilation and quantify volume change, which can be a useful diagnostic aid. However, lung registration is a challenging problem because of the variation in deformation along the lungs, sliding motion of the lungs along the ribs, and change indensity. Landmark correspondences have been used to make deformable image registration robust to largedisplacements. To tackle the challenging task of intra-patient lung computed tomography (CT) registration, we extend the landmark correspondence prediction model deep convolutional neural network-Match by introducing a soft mask loss term to encourage landmark correspondences in specific regions and avoid the use of a mask during inference. To produce realistic deformations to train the landmark correspondence model, we use data-driven synthetic transformations. We study the influence of these learned landmark correspondences on lung CT registration by integrating them into intensity-based registration as a distance-basedpenalty. Our results on the public thoracic CT dataset COPDgene show that using learned landmark correspondences as a soft constraint can reduce median registration error from approximately 5.46 to 4.08mm compared to standard intensity-based registration, in the absence of lungmasks. We show that using landmark correspondences results in minor improvements in local alignment, while significantly improving globalalignment.

Prediction and decision of free-piston linear generator on starting process for multi-fuel adaptability

Free-piston linear generators, due to their variable compression ratio, are considered to possess excellent multi-fuel adaptability. Further, fuel-flexible is a weighty research topic for addressing energy source limitations, meeting powertrain generalization requirements, and achieving energy savings and emission reductions. The initial concern to solve is the rapid starting of various fuels within the same free-piston linear generator. This study aims to introduce a mechanical resonance starting performance prediction model based on an artificial neural network and provides a control decision method based on single-objective optimization with soft constraints to achieve flexible starting of a variety of fuels. This paper also assessed the starting performance using the free-piston linear generator prototype and conducted a comprehensive parameterized analysis to investigate the influence of parameters and control variables on starting performance. Using the starting decision method proposed, the control variable values for ethanol starting were derived on the prototype, resulting in driving current at 3.5A, intake air mass flow at 95 standard liters per minute, and intake air temperature at 335 K. This decision value effectively met the predefined requirements, successfully igniting and starting the prototype with ethanol. This outcome serves as a robust validation of the fuel-flexible capabilities within free-piston linear generator systems.

Solving forward and inverse problems of contact mechanics using physics-informed neural networks

This paper explores the ability of physics-informed neural networks (PINNs) to solve forward and inverse problems of contact mechanics for small deformation elasticity. We deploy PINNs in a mixed-variable formulation enhanced by output transformation to enforce Dirichlet and Neumann boundary conditions as hard constraints. Inequality constraints of contact problems, namely Karush–Kuhn–Tucker (KKT) type conditions, are enforced as soft constraints by incorporating them into the loss function during network training. To formulate the loss function contribution of KKT constraints, existing approaches applied to elastoplasticity problems are investigated and we explore a nonlinear complementarity problem (NCP) function, namely Fischer–Burmeister, which possesses advantageous characteristics in terms of optimization. Based on the Hertzian contact problem, we show that PINNs can serve as pure partial differential equation (PDE) solver, as data-enhanced forward model, as inverse solver for parameter identification, and as fast-to-evaluate surrogate model. Furthermore, we demonstrate the importance of choosing proper hyperparameters, e.g. loss weights, and a combination of Adam and L-BFGS-B optimizers aiming for better results in terms of accuracy and training time.

Corr-Track: Category-Level 6D Pose Tracking with Soft-Correspondence Matrix Estimation.

Category-level pose tracking methods can continuously track the pose of objects without requiring any prior knowledge of the specific shape of the tracked instance. This makes them advantageous in augmented reality and virtual reality applications. The key challenge is how to train neural networks to accurately predict the poses of objects they have never seen before and exhibit strong generalization performance. We propose a novel category-level 6D pose tracking method Corr-Track, which is capable of accurately tracking objects belonging to the same category from depth video streams. Our approach utilizes direct soft correspondence constraints to train a neural network, which estimates bidirectional soft correspondences between sparsely sampled point clouds of objects in two frames. We first introduce a soft correspondence matrix for pose tracking tasks and establish effective constraints through direct spatial point-to-point correspondence representations in the sparse point cloud correspondence matrix. We propose the "point cloud expansion" strategy to address the "point cloud shrinkage" problem resulting from soft correspondences. This strategy ensures that the corresponding point cloud accurately reproduces the shape of the target point cloud, leading to precise pose tracking results. We evaluated our approach on the NOCS-REAL275 and Wild6D dataset and observed superior performance compared to previous methods. Additionally, we conducted cross-category experiments that further demonstrated its generalization capability.

Enhancing human–robot collaborative transportation through obstacle-aware vibrotactile warning and virtual fixtures

Transporting large and heavy objects can benefit from Human–Robot Collaboration (HRC), increasing the contribution of robots to our daily tasks and addressing challenges arising from labor shortages. This strategy typically positions the human collaborator as the leader, with the robot assuming the follower role. However, when transporting large objects, the operator’s situational awareness can be compromised as the objects may occlude different parts of the environment, weakening the human leader’s decision-making capacity and leading to failure due to collision. This paper proposes a situational awareness framework for collaborative transportation to face this challenge. The framework integrates a multi-modal haptic-based Obstacle Feedback Module with two units. The first unit consists of a warning module that alerts the operator through a haptic belt with four vibrotactile devices that provide feedback about the location and proximity of the obstacles. The second unit implements virtual fixtures as hard constraints for mobility. The warning feedback and the virtual fixtures act online based on the information given by two Lidars mounted on a mobile manipulator to detect the obstacles in the surroundings. By enhancing the operator’s awareness of the environment, the proposed module improves the safety of the human–robot team in collaborative transportation scenarios by preventing collisions. Experiments with 16 non-expert subjects in four feedback modalities during four scenarios report an objective evaluation thanks to quantitative metrics and subjective evaluations based on user-level experiences. The results reveal the strengths and weaknesses of the implemented feedback modalities while providing solid evidence of the increased situational awareness of the operator when the two haptic units are employed.

Automated Course Timetabling Optimization Using Tabu-Simulated Annealing Hyper-Heuristics Algorithm

The topic of solving Timetabling Problems is an interesting area of study. These problems are commonly encountered in many institutions, particularly in the educational sector, including universities. One of the challenges faced by universities is the Course Timetabling Problem, which needs to be addressed regularly in every semester, taking into consideration the available resources. Solving this problem requires a significant amount of time and resources to create the optimal schedule that adheres to the predefined constraints, including both hard and soft constraints. As a problem of computational complexity, University Course Timetabling is NP-hard, meaning that there are no exact conventional algorithms that can solve it in polynomial time. Several methods and algorithms have been proposed to optimize course timetabling in order to achieve the optimal results. In this study, a new hybrid algorithm based on Hyper-Heuristics is developed to solve the course timetabling problem using the Socha Dataset. This algorithm combines the strengths of Simulated Annealing and Tabu Search to balance the exploitation and exploration phases and streamline the search process. The results show that the developed algorithm is competitive, ranking second out of ten previous algorithms, and finding the best solution in six datasets.

Path Optimization of Container Multimodal Transportation Considering Differences in Cargo Time Sensitivity

To optimize the container multimodal transportation path selection problem with differences in cargo time sensitivity, this paper introduces the concept of mixed time windows and establishes an integer programming model from the perspective of carriers. The objective is to minimize the total transportation cost while meeting the customer’s on-time delivery requirements. The model imposes hard time window constraints on cargoes with on-time delivery requirements and soft time window constraints on general cargoes. A bi-level genetic algorithm is employed to solve the model. A case study is conducted to demonstrate the effectiveness of the proposed approach. Experimental results show that the bi-level genetic algorithm’s convergence ability, efficiency, stability, and global search capabilities are far superior to those of traditional single-level genetic algorithms. This indicates that the use of multi-stage solution approaches can mitigate the shortcomings of genetic algorithms in dealing with high-dimensional problems. In addition, when analyzing the effects of time penalty intensity and time window on decision-making results, it is found that the more time-sensitive cargoes are more likely to be affected by the time window setting, and will tend to sacrifice costs to ensure that the time demand is met. At the same time an increase in demand for time-sensitive services is accompanied by an increase in operating costs. Accurate assessment and classification of time sensitivity of cargoes can effectively prevent wasting resources in unnecessary areas, and ultimately maximize decision outcomes and ensure the most efficient use of resources.

Bundle adjustment with motion constraints for uncalibrated multi-camera systems at the ground level

Multi-camera systems for structure from motion (SfM) are widely deployed in many mapping applications. Existing solutions assume known rig calibration, synchronized frames among cameras, as well as overlapping field of views (FoVs). In this paper, we derive novel geometric constraints assuming minimal knowns about the multi-camera systems, to benefit low-cost and non-expert use cases where uncalibrated multi-camera systems with non-typical geometry setups present, i.e., no rig calibration, no overlapping FoVs. Assuming that these cameras are co-located and share the same motion of the platform, the proposed constraints utilize the parallelism and length proportionality of motion vectors of these co-located cameras and formulate them as translation constraints into the bundle adjustment (BA). The proposed constraints (called motion constraints) impose a first-order penalty to co-located cameras whose motion speeds and directions between frames do not match. With soft constraints, it can handle loosely synchronized frames (with an error within one second). The proposed constraints are integrated into the BA framework and experimented with different camera setups, i.e., on a group of casually co-located GoPro cameras with no rig calibration, and some with no overlapping views. Our results show that the constraints are extremely effective in improving the reconstruction and pose accuracy for ground motion images: in our self-collected open trajectories without loop closure, the proposed constraints are effective in correcting topographical errors (i.e., trajectory drifts) of the resulting models, and the dense point clouds achieve up to 11.34 m (86.12 %) of mean absolute error (MAE) improvement as compared to reference LiDAR point clouds; our results on KITTI-odometry and KITTI-360 datasets also show an improvement of up to 28.82 m (81.05 %) in terms of the root mean square error (RMSE) of absolute pose error (APE). We expect that the proposed constraints are significant not only as additional geometric constraints for image-based mobile mapping, but also will benefit the broader use of photogrammetry, since it empowers the possibility to harness the traditionally so-called low-quality stereo/multi-camera data (e.g., by non-photogrammetry citizen scientists) into improved 3D products.

A new algorithm for computing the nearest polynomial to multiple given polynomials via weighted ℓ2,q-norm minimization and its complex extension

A new algorithm is proposed in this paper for computing the nearest polynomial to multiple given polynomials with a given zero in the real case, where the distance between polynomials is defined by the weighted ℓ2,q norm (0<q≤2). First, the problem is formulated as a univariate constrained non-convex minimization problem, where prior information of the coefficients of the polynomial can be embedded by selecting proper weights. Then, an iteratively reweighted algorithm is designed to solve the obtained problem, and also the convergence and rate of convergence are uniformly demonstrated for all q in (0,2]. Since all the existing methods for computing the nearest polynomial to multiple given polynomials with a given zero are limited to the real case, we ingeniously extend the results to the complex case. Finally, two representative examples that separately compute the nearest real and complex polynomials are presented to show the effectiveness of the proposed algorithm.

Research on a Simulation and Prediction Method of Traffic Flow Information Based on PINN

Traffic flow prediction can help intelligent transportation effectively be carried out. This paper employs the Physics-Informed Neural Network (PINN) as a deep learning method and partial differential equation (PDE) multivariable methods to approximate solutions for the Lighthill-Whitham-Richards (LWR) traffic flow equation system. The relationship between traffic flow density function and time/displacement is obtained by setting different function conditions and initial boundary constraints. Furthermore, a series of three-dimensional graphics are generated during the research process through dynamic comparisons and evaluations of independent variables to intuitively display and analyze traffic flow trends and gradients. Additionally, before the commencement of the study, the authors propose specific initial condition combinations corresponding to real-world application scenarios to enhance the practicality of the research. Moreover, through the interpretation and reflection on the obtained images, the study identifies inconsistencies between actual situations and predictions, leading to necessary adjustments. Furthermore, the research explores the compatibility of the u(x) function and hard constraint conditions in different scenarios, aiming to accurately predict traffic information and ultimately enhance the utilization of traffic resources. By leveraging PINN and PDE technologies, this study contributes to advancing traffic flow analysis and optimization, potentially influencing urban planning and traffic management strategies positively.

A six degrees-of-freedom cable-driven robotic platform for head-neck movement.

This paper introduces a novel cable-driven robotic platform that enables six degrees-of-freedom (DoF) natural head-neck movements. Poor postural control of the head-neck can be a debilitating symptom of neurological disorders such as amyotrophic lateral sclerosis and cerebral palsy. Current treatments using static neck collars are inadequate, and there is a need to develop new devices to empower movements and facilitate physical rehabilitation of the head-neck. State-of-the-art neck exoskeletons using lower DoF mechanisms with rigid linkages are limited by their hard motion constraints imposed on head-neck movements. By contrast, the cable-driven robot presented in this paper does not constrain motion and enables wide-range, 6-DoF control of the head-neck. We present the mechatronic design, validation, and control implementations of this robot, as well as a human experiment to demonstrate a potential use case of this versatile robot for rehabilitation. Participants were engaged in a target reaching task while the robot applied both assistive and resistive moments on the head during the task. Our results show that neck muscle activation increased by 19% when moving the head against resistance and decreased by 28-43% when assisted by the robot. Overall, these results provide a scientific justification for further research in enabling movement and identifying personalized rehabilitation for motor training. Beyond rehabilitation, other applications such as applying force perturbations on the head to study sensory integration and applying traction to achieve pain relief may benefit from the innovation of this robotic platform which is capable of applying controlled 6-DoF forces/moments on the head.

Research on two stage obstacle-avoidance trajectory planning and trajectory tracking control in curves

The existing obstacle-avoidance trajectory planning and trajectory tracking control algorithms have limitations such as long-time consumption, high failure rate in dynamic traffic environments, and insufficient trajectory tracking accuracy in curved roads. Based on the above problems, this paper designs a two stage obstacle-avoidance trajectory planner based on nonlinear optimization theory. In first stage Part-NLP, only considering the safety obstacle avoidance, a point mass model and linearization constraints are established to quickly solve the initial trajectory. In the second stage Full-NLP, considering smooth soft constraints comprehensively, the initial trajectory is optimized by establishing driving corridors and a lightweight iterative framework. In control module, this paper selects a linear quadratic form lateral trajectory tracking controller, and the parameters were optimized through the carnivorous plant algorithm. The joint simulation results show that in dynamic traffic environment of curved roads, the two stage planner proposed can accurately plan safe and smooth obstacle avoidance trajectories, and there is a significant reduction in time consumption compared to traditional NLP algorithms. The control strategy can accurately track the planned trajectories, with lateral error controlled within plus or minus 0.1 m, heading error controlled within plus or minus 0.15 rad, speed tracking error controlled within plus or minus 0.15 m/s, and vehicle yaw angle error controlled within plus or minus 0.04 rad; the hardware-in-loop test results indicate that the controller can achieve real-time and accurate trajectory tracking.

Regulating local government debt in China: Intended and unintended consequences

This study integrates institutional and empirical analyses to investigate the intricate policies that regulate local government debt in China. It finds that, within a remarkably short period, China has skilfully incorporated local government debt into the local budget management system, thereby transforming it from a mere transaction between local governments and credit agencies to an intergovernmental fiscal relations framework. The study also establishes that the pivotal debt-regulating policies are more geared towards prioritizing the cost of debt over its scale. Official documents, statistical data, and in-depth interviews provide evidence to support these findings. Drawing on an analysis of county-level panel data in Sichuan Province, this study identifies two main effects of these policies. Primarily, the cost of local general debt may trigger the notorious soft budget constraint, thereby incentivizing county-level governments to depend more heavily on superior transfer payments. Additionally, the fiscal consolidation policy, which focuses on the cost of local special debt, may lower the debt ratio. However, it also has the unintended consequence of bolstering the local government’s reliance on land finance.

Database Repairing with Soft Functional Dependencies

A common interpretation of soft constraints penalizes the database for every violation of every constraint, where the penalty is the cost (weight) of the constraint. A computational challenge is that of finding an optimal subset: a collection of database tuples that minimizes the total penalty when each tuple has a cost of being excluded. When the constraints are strict (i.e., have an infinite cost), this subset is a “cardinality repair” of an inconsistent database; in soft interpretations, this subset corresponds to a “most probable world” of a probabilistic database, a “most likely intention” of a probabilistic unclean database, and so on. Within the class of functional dependencies, the complexity of finding a cardinality repair is thoroughly understood. Yet, very little is known about the complexity of finding an optimal subset for the more general soft semantics. The work described in this manuscript makes significant progress in that direction. In addition to general insights about the hardness and approximability of the problem, we present algorithms for two special cases (and some generalizations thereof): a single functional dependency, and a bipartite matching. The latter is the problem of finding an optimal “almost matching” of a bipartite graph where a penalty is paid for every lost edge and every violation of monogamy. For these special cases, we also investigate the complexity of additional computational tasks that arise when the soft constraints are used as a means to represent a probabilistic database in the case of a probabilistic unclean database.