Posters & Demonstrations

Attitude Controllers are commonly used in aerospace applications to rotate a spacecraft into a desired orientation. However, when the goal is simply to align a body-frame vector with an inertial vector, this is inefficient because time and propellant are wasted aligning the roll angle of the body about the aligned target vector, which is unnecessary. Instead, the Pointing Controller’s goal is simply the single vector alignment with zero angular velocity. The aim of this thesis research will be to derive attitude thruster-controlled trajectories for this Pointing Controller, optimized for minimum time using Sequential Convex Optimization. Various reference trajectories, such as 1DOF planar and unstable constant torque bang-bang control, are explored to minimize both time of flight and computation. Depending on Monte Carlo tests of various trajectories, this may yield guidance techniques that are fast enough to be practical for onboard systems, and thus very likely the optimal technique for spacecraft vector pointing control.

Disasters such as hurricanes and floods increasingly disrupt human access to food, medicine, and emergency supplies. Drones have been considered as promising tools for emergency supplies delivering; however, their effectiveness is constrained by limited battery life. Moreover, extreme environment during disasters makes it hard to recharge drone batteries, while the duration of most delivery missions is uncertain. Therefore, this research develops a self-charging drone equipped with a Hybrid Energy Collecting and Storage System to extend battery life and ensure sufficient power during operation. Specifically, Hybrid Energy Collecting and Storage System is the power management system that absorbs energy from environment, such as wind, solar, and thermal energy. The harvested energy is stored in a supercapacitor and released to recharge the battery whenever its capacity falls below 30%. This project utilizes Python programming and real meteorological data from the National Centers for Environmental Information (NCEI) for Tampa, Florida, collected after Milton Hurricane, to stimulate the performance of the Hybrid Energy Collecting and Storage System. Beyond energy control, the project employs the Robot Operating System (ROS) to generate a 3-D environmental maps using LiDAR and GPS data. A mapping coordinator node identifies delivery zones and computes efficient flight trajectories while accounting for real-time energy availability. The integration of ROS and Python enables the drone to map delivery zone autonomously, optimize routes for minimal energy cost, and self-charge. Experiments and analyses show that under realistic post-hurricane weather conditions, the Hybrid Energy Collecting and Storage System can extend flight endurance by more than 35% compared to conventional battery-powered UAVs, maintaining operation even under low sunlight or weak wind conditions.

Recent advances in manufacturing have enabled the rapid prototyping of 3D-printed compliant robotic grippers, increasing the focus of researchers on optimizing their structure to improve performance. The ability of these grippers is greatly influenced by both the infill geometry and density, which together govern the force transmission and compliance of the gripper. Compliance is essential because it allows the gripper to conform to an object’s shape, enabling more uniform distribution of contact forces. At the same time, effective force transmission is required to sufficiently grasp the object. Together these properties enable the manipulation of delicate or irregularly shaped objects without breaking them. Prior work completed in the lab has shown that finite element analysis (FEA) can accurately predict real-world gripper performance by evaluating maximum point displacement and stress under prescribed loads. While faster than manually testing grippers, running FEA on a large set of both infill patterns and densities still remains largely time consuming and manual. The inability of researchers to explore the vast design space of possible grippers limits the potential to optimize gripper structure and function.

In this work, we present a tool that automates the generation, simulation, and comparison of diverse infill architectures for predefined geometries. The software procedurally generates various common 3D-printing infills, including fin-ray, grid, honeycomb, and triangle, at differing densities, generates a mesh, and runs a predefined FEA pipeline to calculate the stress and displacement for a given load. Then, using a formula developed in the lab that combines both metrics, the software can rank gripper structures, providing researchers with valuable information on the viability of any infill for their specific use-case. Preliminary results have demonstrated agreement between rankings determined by the software, manual FEA, and real-world testing. Ultimately, by improving the ability of researchers and manufacturers to find optimal gripper infills, the proposed tool can be used to create better products for a wide array of application cases.

This project involves developing an automated workflow for preparing large power-system datasets for dynamics simulation. Raw data of the WECC-240 bus system, provided by NREL (National Renewable Energy Laboratory), is distributed across two different files - the .raw file containing the static machine data, and the .dyr file containing the dynamic data - that detail information for buses, branches, transformers, loads, generators and IBRs. The presidential files are extremely dense, making manual parsing very time consuming and inefficient. Thus, an initial semi-manual process was adopted to understand the processing logic before automation. This involved extracting specific parts of the raw data relevant to the system component, identifying and assigning the data to the variables involved, and processing them before producing the formatted output files using Python. As the project matured this process was fully automated, resulting in a consistent and reproducible pipeline capable of handling large input datasets with minimal user intervention.

The automated system now generates standardized .m and .py files that serve as the input to a pre-existing code that solves the power system dynamics. This reduces human error in inputting data, and also speeds up the process. To improve usability, a custom user interface (UI) was developed to streamline the entire workflow, allowing users to upload raw data files, run the full processing pipeline, and receive the organized output files without interacting with the back-end scripts. The tool also produces a unified JSON file from the raw data, creating a coherent, searchable data structure.

The Data Explorer component of the UI displays the JSON file in a clean, human-readable grid. Users can navigate through complex multi-field system data, filter values, verify consistency, and inspect specific components without switching tools, allowing for a smoother debugging phase. In addition to data inspection, the software includes an in-built validation check for the raw data, as well as visualization capabilities for the WECC-240 bus system using spring, Kamada-Kawai, and circular layouts. These graphical views help users confirm network connectivity, detect anomalies, and better understand the structure of the grid.

Ongoing development involves integration of the pre-existing simulation code with the GUI. This enables users to run dynamic simulations directly through the UI and view outputs such as eigenvalue plots, stability metrics, and control-model behavior. Together, visualization enhancements, work-flow improvements in the automation and organization of data, and direct integration with the code improves accessibility, accuracy and efficiency for researchers working with large power-system datasets.

By: Mahrukh Anindya, Ruhaan Batta, Trinay Ravella

UAV path planning in cluttered environments becomes especially challenging when wind conditions vary sharply around obstacles, creating nonlinear flow patterns that cannot be captured by uniform wind assumptions used in many existing planners. This work introduces CFDMPC, a framework that integrates physics based wind modeling with nonlinear model predictive control (NMPC) to enable safe navigation through such environments. We leverage a physics informed neural network (PINN) to solve the governing partial differential equations (PDEs), specifically the steady state Reynolds Averaged Navier Stokes (RANS) equations, without requiring any training data. By embedding CFD physics directly into the loss function, the PINN rapidly predicts obstacle induced wind fields at runtime, avoiding the high computational cost of full CFD simulations. These spatially varying wind estimates are then incorporated into the MPC prediction model, allowing the UAV to anticipate local aerodynamic disturbances and adjust its trajectory accordingly. Experiments across environments with one, two, and four obstacles demonstrate the advantages of the proposed approach. CFDMPC consistently reaches the goal without collisions, whereas a constant wind baseline often fails in regions with strong velocity gradients. Moreover, CFDMPC improves efficiency; in the two obstacle scenario, it reduces total control effort by approximately 9 percent relative to a CFD driven MPC planner. Although trajectory length and computation time naturally increase with obstacle density, CFDMPC remains robust and reliable across all scenarios. These results highlight that coupling fast, PDE based wind inference with real time MPC significantly improves UAV safety, predictability, and energy efficiency in realistic, obstacle disturbed wind environments.

This research develops an autonomous navigation system for maritime environments, addressing the dynamic challenges of aquatic navigation. Unlike land-based autonomous systems that navigate relatively static environments, our approach must account for shifting wave patterns, unpredictable currents, and transient obstacles. Navigation is conducted using relatively few landmarks present on otherwise flat water. Using an integrated multi-sensor localization method, we aim to achieve real-time obstacle detection, localization, and adaptive pathfinding in complex water conditions. Our model incorporates data from four RGB depth cameras for identification of markers and inertial measurement units. Sensor inputs are processed through a pipeline designed to filter out noise from water movement and sensor error, ensure compatibility between sensor systems that output data at different rates, and prioritize critical real-time data to path-find its destination. Though results are pending, this technology has significant applications across global trade, disaster response, defense, and environmental monitoring. By advancing autonomous navigation at sea, this project aims to reduce human risk and improve operational efficiency.

Reinforcement learning has recently enabled promising progress in humanoid loco-manipulation. However, most policies operate at low frequency and rely on PD control for torque tracking. Consequently, the policy must produce both feedforward and feedback components through a single control channel. In contrast, model-based controllers provide precise planning via online optimization but are sensitive to modeling errors. Attempts to combine both approaches face competing demands between GPU-parallel RL for high sample efficiency and the iterative constrained optimizers used in model-based control.

We introduce a centroidal whole-body controller that serves as a drop-in replacement for PD control in standard RL pipelines, splitting control between a feedforward centroidal controller and feedback PD controller. Additionally, our formulation simplifies model-based control by using a learned contact state and reference torques, shifting constraints previously enforced through optimization to constraints implicitly enforced by the RL policy, allowing for fast, closed-form computation of joint torques.

We validate the combined RL–centroidal framework on multiple loco-manipulation tasks using existing motion-retargeting pipelines. Our experiments demonstrate up to a 22% reduction in base-velocity tracking error for a balancing task with only an 11% increase in computational cost, showing that model-based control can be integrated into RL pipelines efficiently.

Digital Glass Forming (DGF) is a manufacturing process that uses a laser to locally heat glass, creating a workable volume of material that can be shaped. When a filament or fiber is continuously fed into the work zone, the process enables additive manufacturing of glass components. While it is relatively straightforward to fabricate consistent morphologies once the process reaches steady state, achieving repeatable and reliable morphologies during unsteady periods is significantly more challenging. Examples of unsteady deposition include starting and ending a track, sudden changes in deposition direction, and closing a connected track. In this poster, we discuss our work on modeling unsteady glass deposition and the control strategies we have developed to achieve precise deposition using a visual camera. Our real-time control system operates on Linux and coordinates a fiber laser, a filament feeder, and a motion system upon which the part is fixed. A thermal camera, visual camera, and confocal displacement sensor are used for real-time data acquisition, and all data is temporally and spatially registered. In this work, we focus on the specific unsteady fabrication case of deposition starts. To achieve consistent starts, we heat the filament to produce a repeatable tip size and position, then accelerate it toward the substrate at the desired velocity. As the filament is heated, a molten ball forms at the tip and grows as it moves up the filament. We modeled the height and area of this molten ball as functions of laser power and filament displacement. This physics-based and data-driven model is integrated into a PI controller to regulate the tip size and position by determining the appropriate laser power command profiles. Iterative Learning Control (ILC) will then be applied to determine the filament feed velocity profiles that yield consistent starting and ending profiles such that, as measured by the camera, the track profile matches the steady-state morphology and begins/ends at the desired locations. A set of experiments will be conducted using constant process parameters, and another set will be conducted with closed-loop control.

Human-autonomy teaming (HAT) is becoming critical across many sectors of society, particularly as autonomous systems (enhanced by AI) advance in capability. As a result, it is important for humans to be able to collaborate effectively with robots. Prior research has shown that team cognition is an emergent state that promotes effective team performance, with communication being essential for team success. This is true for both human-human and human-autonomy teams. Communication allows teammates to share information to help maintain situational awareness, cultivate trust, and reduce workload, in turn impacting individual and team performance. Therefore, understanding and modeling the relationship between communication, human cognitive states, and team performance is essential toward advancing the state-of-the-art in human-autonomy teaming, particularly for teams comprised of multiple humans and agents. Critical research questions remain unanswered, such as “how can we design a scalable set of models to inform communication content and frequency between the agents and the human to maximize performance outcomes?” The objective of this research is to build a closed-loop algorithmic framework that optimizes bidirectional communication features to influence human cognitive states and thereby improve team performance.

Many people worldwide live with upper limb disabilities stemming from various neurological conditions, such as spinal cord injuries, multiple sclerosis, and stroke. Their loss of hand function greatly limits their ability to perform activities of daily living (ADLs) such as eating, handling household items, and grooming, which reduces their independence and diminishes their quality of life.

Assistive robotic manipulators, particularly wheelchair-mounted robotic arms (WMRAs), have been adopted to aid the user in better completing ADLs independently. However, current commercial systems rely on rigid grippers, which are suited for precise controlled tasks, unlike inherently variable ADLs. This lack of flexibility imposes immense cognitive load on the user, who depends solely on visual judgment to stably manipulate the object without dropping or breaking it. Alternatively, compliant grippers are being investigated as they passively conform to object surfaces, better distributing the applied force along them and enabling better manipulation of both delicate and irregularly shaped objects that may be involved in ADLs, such as food items.

In this study, ten human subjects were recruited to experimentally judge and control the grasp of both compliant and rigid grippers in a WMRA-like setup. They were tasked with modulating the grasp on ADL-related objects of varying geometries and fragilities without dropping or breaking them from a seated position, using both gripper types and only their visual judgement. Their comparison was assessed via grasp adjustment times, end positions of the gripper fingers, success/failure rates, and System Usability Scale scores with the compliant grippers demonstrating better overall performance, supporting their potential adoption in assistive robotic manipulation to enhance user independence.

Digital twins provide a framework for extracting data from control systems and sensors to predict real-time process behavior, enabling process optimization, operational efficiency, and safety management. Traditional methods for digital twin development such as trial-and-error and statistically grounded design of experiments have been successful but face limitations for high-dimensional, noisy dynamic systems. Model-based design of experiments (MBDoE) offers a more rigorous alternative by identifying experiments that yield maximal information about unknown parameters, thereby enabling uncertainty quantification. Pyomo.DoE is an open-source Python package within the Pyomo environment that facilitates MBDoE using Fisher information matrix (FIM)-based statistical inference techniques. FIM-based design criteria such as D-, A-, E-, and ME-optimality form the basis of MBDoE within Pyomo.DoE. To generate approximations of the FIM, computation of the dynamic sensitivity (i.e., the Jacobian) matrix is critical. Two strategies exist for computing Jacobians: (1) finite difference (numerical derivative) schemes and (2) direct sensitivity equation methods. The current Pyomo.DoE implementation uses a central finite difference formula with 2n+1 parameter perturbations (for n unknown parameters) within a two-stage optimization structure. This approximation introduces numerical errors and results in massive optimization problems for larger systems. In contrast, commercially available tools such as gPROMS integrate sensitivity equations and/or leverage automatic differentiation to obtain first derivatives to compute the FIM. This work introduces an enhancement to the Pyomo.DoE framework that computes Jacobians analytically via the automatic differentiation capabilities in PyNumero, Pyomo’s optimization backend. Analytical derivatives of the model equations are generated, and an auxiliary system is solved to obtain the Jacobian, reducing numerical error and computational time. Unlike classical sensitivity-based methods, this approach is generalizable to any well-posed Pyomo model and is not restricted to dynamic systems. The enhanced version is benchmarked against the central finite difference method using dynamic systems ranging from single-parameter ODEs to multi-parameter PDE systems. Results show that analytical differentiation reduces computation time by a factor of two, with greater savings for more complex models. These findings indicate that the enhanced Pyomo.DoE framework can accelerate MBDoE and facilitate its integration into digital twin design for real-time processes.

Autonomous landing systems seek to maximize the reusability and sustainability of launch vehicles. The Collegiate Lander Challenge invites student teams from across the country to complete a series of challenges in developing an autonomously landing rocket. ASTRA is a vertical-landing rocket testbed developed by Purdue Space Program's Active Controls (AC) Team that seeks to test GNC and Avionics systems on a low-stakes, EDF-propelled rocket before building up to a full autonomously landing liquid rocket. The vehicle is fitted with onboard inertial sensing and actuators capable of modulating thrust and attitude, which enables real-time closed-loop control. Purdue Space Program's Astrodynamics and Space Applications (ASA) Team is working in collaboration with AC to develop a high level trajectory optimization layer using sequential convex programming (SCP) that computes dynamically feasible, fuel-efficient landing trajectories, and an onboard MPC layer, using a similar SCP formulation, that tracks those trajectories while rejecting disturbances that may arise. The online MPC system will run on ASTRA's embedded flight computer and is designed to be resource efficient. To verify this, hardware-in-the-loop tests are performed. Initial simulations show that this trajectory optimization and MPC framework can consistently guide the vehicle from perturbed initial conditions to a safe, upright landing while satisfying thrust and attitude constraints. Further, Monte Carlo campaigns confirm that the controller maintains small enough terminal position and attitude errors and respects actuator limits. This work contributes a reusable experimental framework for optimal control of landing rockets that enables future iterations to build on its design. Next steps involve integrating both the trajectory optimization and MPC algorithms with ASTRA for flight testing. Beyond enabling future PSP flight tests, these algorithms and ASTRA provide a platform that the broader autonomy and control community at Purdue can explore such as robust MPC, real-time control, and resilient aerospace systems.

Managing misalignment between the supply and demand of power on the electrical grid is becoming increasingly important with the growing reliance on renewable energy produced from sources such as solar and wind. Building HVAC systems play an important role on this system given that they constitute a substantial electrical load on the grid. Thermal energy storage (TES) can help manage this misalignment through load shifting energy used to power HVAC systems. Compared against traditional methods of sensible heat storage, using a composite phase change material-based TES can allow for improved heat transfer and energy density. To best utilize this phase change material-based TES, an accurate estimate for the state of charge is required. Dynamic modeling and state estimation theory can be applied to achieve an improved state of charge estimate. There also exists a need for optimal placement of sensors within the TES, both to improve performance and to minimize the number of required sensors, which can aid in the commercial viability of such a system. This work proposes a method for determining optimal sensor placement within a TES device using dynamic modeling and estimation co-design.

We present a robust safety-critical control framework tailored for networked susceptible-infected-recovered (SIR) epidemic dynamics, leveraging control barrier functions (CBFs) and robust control barrier functions to address the challenges of epidemic spread and mitigation. In our networked SIR model, each node must keep its infection level below a critical threshold, despite dynamic interactions with neighboring nodes and inherent uncertainties in the epidemic parameters and measurement errors, to ensure public health safety. We first derive a CBF-based controller that guarantees infection thresholds are not exceeded in the nominal case. We enhance the framework to handle realistic epidemic scenarios under uncertainties by incorporating compensation terms that reinforce safety against uncertainties: an independent method with constant bounds for uniform uncertainty, and a novel approach that scales with the state to capture increased relative noise in early or suppressed outbreak stages. Simulation results on a networked SIR system illustrate that the nominal CBF controller maintains safety under low uncertainty, while the robust approaches provide formal safety guarantees under higher uncertainties; in particular, the novel method employs more conservative control efforts to provide larger safety margins, whereas the independent approach optimizes resource allocation by allowing infection levels to approach the boundaries in steady epidemic regimes.

Tactile feedback provides rich contact information for robotic actuators, offering detailed insights into object shape, texture, contact forces, deformation, and slip. Vision-based tactile sensors widely used in robotic systems typically employ a deformable gel or membrane that may be coated with fiducial markers or left unmarked, enabling operation in either a marker-based or markerless mode. However, this design decision restricts the sensor to a single operating mode, limiting its versatility: marker-based sensing excels in precise manipulation, while markerless sensing is preferred for broader perception tasks.

Yet, fiducial markers—though indispensable for accurate contact geometry and motion tracking—also occlude the camera and degrade perception quality. This trade-off forces practitioners to choose between manipulation-oriented or perception-oriented sensing, creating a fundamental bottleneck in tactile robotics. To address this challenge, we propose a novel mode-switchable optical tactile sensing framework that enables seamless transitions between marker and markerless modes. The marker-to-markerless transformation is achieved using a generative model, while the inverse conversion is realized through a sparsely supervised regression model. Our approach allows a single-mode tactile sensor to operate effectively in both regimes without requiring any additional hardware modifications. This capability unlocks a unified tactile sensing modality suitable for both perception and manipulation, significantly broadening the operational flexibility of optical tactile sensors in robotic applications.

Our task is to design controllers for autonomous racing's racing cars. We hope that the racing cars can still drive quickly, stably and safely without the intervention of human instructions.

As the team most responsible for the controller design, we need to design a system that converts the best driving path desired by the racing car into actual operation instructions, so that the physical components of the racing car such as the engine, steering gear, and suspension can execute the path planning instructions.

In terms of methodology, we are implementing the PID (Proportional-Integral-Derivative) controller and ADRC (active disturbance rejection controller) in both lateral and longitudinal dimensions based on the physical model of a specific autonomous vehicle on MATLAB.

We then use plot to show the tracking, states, yaw, steering and longitudinal acceleration. With the adjustments of the parameters, we lead to better performance in simulation. We will use the Simscape Vehicle library to do the simulation in the future as the visual representation.

Our results show that our PID controller is functional in moderate initial errors. After the PI for speed and PID for lateral, overshoot is initially high but has a short settling time, while saturation constraints are respected. Operation outside of the specific offset range causes the closed-loop to become unstable and diverge from the reference. ADRC will aim to make the controller work within a larger range and still converge. The goal of this research is to make autonomous vehicles run with robustness and accuracy at high speeds. We will continue to conduct tests and develop control methods in the future, attempting to make this autonomously controlled racing car run efficiently and quickly.

We present a data-driven framework for learning adaptable reduced-order dynamics models of nonlinear systems. Our approach combines a nonlinear encoder-decoder architecture with linear latent dynamics, drawing inspiration from Koopman operator theory to embed complex nonlinear dynamics into a low-dimensional space where state evolution is approximately linear. A key feature of our method is that the linear latent structure enables efficient online adaptation using recursive least squares (RLS), allowing the model to refine its dynamics estimates from streaming data without retraining the neural network components. The encoder and decoder are trained offline to learn a suitable latent representation, while the linear dynamics matrices can be updated in real-time as the system operates. We demonstrate our approach on F-16 fighter aircraft simulation data, which exhibits strongly nonlinear and coupled aerodynamics. The learned representations maintain predictive consistency across multi-step rollout horizons, and preliminary experiments on a planar quadrotor show that latent-space RLS can track changing dynamics online. Our results suggest that structured latent models offer a promising architecture for adaptive model-based control, combining the expressiveness of deep learning with the online adaptability of classical estimation.

—This work presents a consensus-based Bayesian framework to detect malicious user behavior in enterprise directory access graphs. By modeling directories as topics and users as agents within a multi-level interaction graph, we simulate access evolution using influence-weighted opinion dynamics. Logical dependencies between users are encoded in dynamic matrices Ci, and directory similarity is captured via a shared influence matrix W . Malicious behavior is injected as cross-component logical perturbations that violate structural norms of strongly connected components (SCCs). We apply theoretical guarantees from opinion dynamics literature to determine topic convergence and detect anomaly via scaled opinion variance. To quantify uncertainty, we introduce a Bayesian anomaly scoring mechanism that evolves over time, using both static and online priors. Simulations over synthetic access graphs validate our method, demonstrating its sensitivity to logical inconsistencies and robustness under dynamic perturbation.

Inverse reinforcement learning is the problem of inferring a reward function from an optimal policy or demonstrations by an expert. In this work, it is assumed that the reward is expressed as a reward machine whose transitions depend on atomic propositions associated with the state of a Markov Decision Process (MDP). Our goal is to identify the true reward machine using finite information. To this end, we first introduce the notion of a prefix tree policy which associates a distribution of actions to each state of the MDP and each attainable finite sequence of atomic propositions. Then, we character an equivalence class of reward machines that can be identified given the prefix tree policy. Finally, we propose a SAT-based algorithm that uses information extracted from the prefix tree policy to solve for a reward machine. It is proved that if the prefix tree policy is known up to a sufficient (but finite) depth, our algorithm recovers the exact reward machine up to the equivalence class. This sufficient depth is derived as a function of the number of MDP states and (an upper bound on) the number of states of the reward machine. These results are further extended to the case where we only have access to demonstrations from an optimal policy. Several examples, including discrete grid and block worlds, a continuous state-space robotic arm, and real data from experiments with mice, are used to demonstrate the effectiveness and generality of the approach.

This paper presents a novel system for flexible automated fabrication of microrobots with embedded permanent magnets, and for the loading of liquid therapeutic drugs and sealing with thermally sensitive wax. Microrobots featuring embedded magnets are more controllable and observable, and are capable of tasks requiring higher forces. In this system, a micromanipulator controls tweezers, and stepper motors actuate a four-stage system that executes different assembly steps. A syringe pump is used to fill drug delivery microrobots, and a wax seal is applied with a brush made from heated copper wires. This brush is capable of efficiently applying an even wax coating to drug delivery robots, sealing the contained therapeutics inside. Vision-based feedback from an overhead microscope camera ensures precise embedded magnet assembly through a combination of image processing algorithms. A single robot takes approximately 192 seconds to assemble, plus 119 seconds for additional embedded magnets beyond the first, corresponding to 45.7% and 51.8% of the time required by a trained human. Drug loading and sealing takes around 165 seconds, 36.7% of the manual operation time, and offers a significant improvement to seal consistency and control of the thickness and area of application. This work advances micro-assembly toward practical medical use by establishing a practical basis for mass production.

Representation learning plays a crucial role in reinforcement learning, especially in complex environments with high-dimensional and unstructured states. Effective representations can enhance the efficiency of learning algorithms by improving sample efficiency and generalization across tasks. This paper considers the Laplacian-based framework for representation learning, where the eigenvectors of the Laplacian matrix of the underlying transition graph are leveraged to encode meaningful features from raw sensory observations of the states. Despite the promising algorithmic advances in this framework, it remains an open question whether the Laplacian-based representations can be learned online and with theoretical guarantees along with policy learning. We address this by formulating an online optimization approach using the Asymmetric Graph Drawing Objective (AGDO) and analyzing its convergence via online projected gradient descent under mild assumptions. Our extensive simulation studies empirically validate the convergence guarantees to the true Laplacian representation. Furthermore, we provide insights into the compatibility of different reinforcement learning algorithms with online representation learning.

Gaussian Process (GP) regression provides a flexible, nonparametric framework for surrogate modeling that naturally quantifies model-form and prediction uncertainty. As summarized in Rasmussen’s foundational tutorial on GP models [1], GPs enable smooth interpolation, closed-form uncertainty estimates, and analytical gradient properties that make them well-suited for optimization under uncertainty. Inspired by the Bayesian calibration philosophy of Kennedy and O’Hagan [2] and more recent developments on Bayesian hybrid modeling for optimization under epistemic uncertainty [3,4,5], this work develops a unified study of embedding GP surrogates into deterministic, robust, and stochastic optimization formulations within Pyomo.

Pyomo provides a general algebraic modeling environment that links statistical surrogate models with large-scale nonlinear programs, allowing GP predictive means and variances to appear directly inside constraints and objective functions. This makes it possible to compare uncertainty-handling philosophies in a consistent setting. Three representative formulations are examined. Deterministic models substitute GP predictive means for expensive mechanistic simulations, yielding rapid optimization suitable for real-time or online applications. Robust formulations incorporate GP posterior variances to build uncertainty sets or chance constraints, allowing explicit control of epistemic uncertainty and reducing constraint violations. Stochastic formulations treat GP-informed uncertainty as a probability distribution, computing expectation-type objectives and risk measures such as CVaR using sparse-grid quadrature or Monte Carlo sampling. These formulations permit systematic evaluation of risk-aware decisions while maintaining computational tractability.

Benchmark chemical-process case studies demonstrate that GP surrogates retain predictive accuracy while enabling efficient uncertainty propagation. Deterministic embeddings provide rapid solutions but may underestimate risk. Robust and stochastic formulations substantially improve reliability. Unlike classical hybrid models, where GPs correct mechanistic-model discrepancy [3], the present framework treats the GP as an independent predictive layer integrated directly with Pyomo’s nonlinear and stochastic programming interfaces. This enables model modularity, solver compatibility (IPOPT/Gurobi), and seamless switching between data-driven and physics-informed components.

Membrane separation processes play an essential role in water treatment, chemical manufacturing, and the recovery of critical materials. However, the design and interpretation of these systems are often hindered by uncertainty in transport models and their associated parameters. This work develops a framework for dynamic modeling and intrusive uncertainty quantification (UQ) to characterize neutral and charged nanofiltration membranes using data from dynamic diafiltration experiments. The approach integrates physics-based modeling and sensitivity analysis within an equation-oriented computational environment.

The framework will employ ParmEst and Pyomo.DoE within the Pyomo ecosystem to perform parameter estimation and experiment design. These intrusive methods directly use the model structure (e.g., derivatives via symbolic/automatic differentiation) for parameter estimation and model discrimination. A weighted sum of squared errors formulation supports robust parameter estimation across heterogeneous datasets. At the same time, Fisher Information Matrix criteria (A-, D-, E-, and ME-optimality) are proposed to guide the selection of operating conditions that maximize information content. These tools facilitate the evaluation of alternative transport models and support the quantitative comparison of multi-ion transport mechanisms.

Experimentally, a dynamic diafiltration apparatus enables automated adjustment of operating conditions, allowing iterative refinement of parameters and experiment designs. This coupling reflects principles central to control and optimization: feedback-driven selection of experimental inputs, sensitivity-guided operation, and efficient allocation of experimental effort. The workflow demonstrates how integrating intrusive UQ with MBDoE strengthens the predictive use of mechanistic transport models while reducing uncertainty in parameter estimates.

In this project, we propose an algorithm for Koopman operator-based neural network modularization of a pre-trained network and its application to policy reuse. Task modularization for a neural network (NN) decomposes a pretrained network into subnetworks and adapts it for the target task. It has been widely studied due to its benefits in improving interpretability and performance on the target task without fine-tuning the network’s parameters. Nevertheless, most existing techniques rely on heuristics without providing a theoretical foundation for the underlying mechanism.

To address this issue, we first propose a decomposition algorithm based on the Koopman operator theory to decompose a pre-trained network into several subnetworks. Considering NN as an autonomous system, we develop Padded Extended Dynamic Mode Decomposition (PEDMD) to approximate an arbitrary NN as a Koopman operator. The Koopman operator computed from PEDMD is then decomposed by applying the Koopman Mode Decomposition (KMD), where each mode represents a distinct subnetwork. Thus, the complex correlation between subnetworks can be represented as a linear operation between Koopman mode-eigenfunction pairs in the Koopman observable space. Furthermore, the original network can be easily reconstructed as a linear combination of subnetworks.

The decomposed modes are further leveraged as experts of a Mixture-of-Experts (MoE) for a new task. Accordingly, one can leverage prior knowledge (modes) from the pre-trained network and adapt to the new inductive bias introduced by the target task. We empirically validate our algorithm with several dynamical systems, showing PEDMD’s performance to decompose and reconstruct a pretrained network, along with the adaptation to a new task using MoE.

(Disclaimer:A shorter version of this paper was presented at, and appears in the proceedings of, IEEE CDC 2025.)

In this paper, we propose an iterative training algorithm for Neural ODEs that provides models resilient to control (parameter) disturbances. The method builds on our earlier work Tuning without Forgetting-and similarly introduces training points sequentially, and updates the parameters on new data within the space of parameters that do not decrease performance on the previously learned training points-with the key difference that, inspired by the concept of flat minima, we solve a minimax problem for a non-convex non-concave functional over an infinite-dimensional control space. We develop a projected gradient descent algorithm on the space of parameters that admits the structure of an infinite-dimensional Banach subspace. We show through simulations that this formulation enables the model to effectively learn new data points and gain robustness against control disturbance.

Open-die forging is a process by which metal is incrementally shaped through repeated strikes from a tool . The process is versatile, able to achieve many material shapes while achieving desirable material properties. While software exists for generating toolpaths for incremental shaping of material, there is currently limited knowledge of how to synthesize real-time controllers for open-die forging processes which take into account the full range of geometric complexity and material property states. Incremental forming necessitates that the material undergoes plastic deformation, which is modeled by nonlinear PDEs on a 3D domain. This is disadvantageous for producing control schemes which are often built upon linear models of the system. Developments in Koopman operator methods present a way to map the nonlinear dynamics of open-die forging to linear dynamics. The Koopman operator has been shown to portray highly nonlinear systems as linear by mapping the nonlinear state-space model to an infinite-dimensional linear model of built on a Hilbert space of measurement functions. Finite-dimensional approximations of this model seek to capture the accuracy of the full infinite-dimensional system using the sparsest collection of measurement functions . By collecting step-by-step (incremental forming operation) data pairs from the system and lifting the data to the higher-dimensional space of sparse measurements, a linear model of the system can be found from the least-squares mapping of the data. Uniaxial deformation simulations run with JAX-FEM in Python will gather data to form a simplified state-space model using Koopman techniques. Performance of the model will be benchmarked against DEFORM, the industry leading software package for forging modeling. This will demonstrate a novel method for producing an accurate, data-driven Koopman model for open-die forging and enable further analysis into the controllability of the open-die forging process and the reachability of material states.

Autonomous drone delivery to door relies on a popular computer vision technique called semantic segmentation (SS) to recognize meaningful house segments and determine a precise drop-off point near the door. While this SS-based approach is effective under specific environments, it fails to perform well across other common environmental factors such as different seasons, hours of the day, weather and illumination levels. In this work, we propose Robust Structural Semantic Segmentation (RSSS), a novel patch to the existing SS solution without requiring re-training for new environments. The core idea is to “Let Strong Help Weak”, where the results of semantic segmentation obtained under favorable/strong conditions are utilized to enhance the weaker ones in adverse settings. This improvement is achieved by leveraging house structures and spatial layouts, which remain largely invariant across various environments. Our evaluation shows that RSSS outperforms the state-of-the-art methods and significantly enhances the robustness of SS and drone delivery across various environments.

Achieving stable humanoid walking on moving platforms with unknown motion remains a challenging control problem due to the hybrid dynamics involving time-varying unknown disturbance. Adaptive control is a widely used method for rejecting such disturbances, yet existing adaptive controllers are typically designed for continuous-time system and do not directly applicable for hybrid systems. This work develops an adaptive controller capable of rejecting multiple sinusoidal disturbances with unknown frequencies, amplitudes, and phases for a linear hybrid system. First, we show that this disturbance rejection problem is solvable given the solution to the hybrid regulator equation exists. Second, we propose an output feedback controller structure and prove that at least one set of controller parameters achieves exact disturbance rejection. Third, using hybrid swapping lemma, we construct a parameter estimation law to adjust the controller parameters using a least-square-type adaptive law. Finally, we treat the humanoid walking on moving surfaces as a case study to validate the usefulness of the proposed method. Both simulation and experiments confirm the performance improvement through an ablation study.

Electric Vehicles (EVs), as the most common type of alternative fuel vehicles, show great potential for decreasing transportation energy consumption and reducing greenhouse gas emissions. In cognizance of these potential benefits, the growing market penetration of EVs, and the anticipated widespread adoption of autonomous vehicles (AVs), it is imperative to improve the energy efficiency of Autonomous Electric Vehicles (AEVs). Previous studies have shown that driver behavior directly affects vehicle energy consumption and automated driving will drastically shift driving behaviors compared to manual driving. As such, the role of autonomous driving technologies in reducing energy consumption is gaining attention. AEVs can potentially enhance driving safety and energy efficiency through high-precision real-time maneuvering. However, compared to gasoline autonomous vehicles (GAVs), research on energy-efficient strategies for AEV operations remain underdeveloped, from the separate perspectives of a single AEV and a network of AEVs. The energy consumption of AEVs is tightly coupled to the autonomous driving behavior and internal energy management but suffers from hurdles including limited battery capacity and slow charging speed. In this context, this paper proposes a novel AEV driving strategy which integrates Vision-Language Models (VLMs) and Reinforcement Learning (RL) to maximize the energy efficiency while preserving driving safety. The methodology involves pretraining VLMs using offline expert driving datasets, thereby enabling them to evaluate safety outcomes as guidance for diverse traffic scenarios. Then, leveraging the safety guidance, a reward framework is developed to generate reward functions that incorporate elements of the surrounding environment: safety, traffic efficiency, and energy efficiency. In sufficiently safe scenarios, the RL-based agent is capable of controlling the speed and driving behavior of ego AEV to minimize energy consumption and maximize safety. To enhance the reliability of proposed driving strategy, an artificial potential field (APF) model was integrated downstream of the RL module. This enabled real-time evaluation of safety based on the state information of surrounding vehicles. The proposed strategy was trained and evaluated through extensive experiments on a CARLA simulation platform. The results demonstrate that the proposed approach effectively reduces energy consumption without affecting the collision rate and route completion rate compared to state-of-the-art (SOTA) baselines.

Tactile perception is significant for robotic operation in unstructured environments. Whisker-based sensors offer lightweight bio-inspired solutions, yet most rely on single-whisker sensing and are limited by passive interaction and constrained functionality. Here we present a circular array of eight independently actuated whiskers, each driven by a pulse-switchable permanent magnet and tracked by a camera. This design enables simultaneous multi-point sensing and coordinated actuation, supporting diverse functions. Quantitative analyses demonstrate accurate pixel-to-physical mapping, consistent pixel-to-force characterization, and long-term repeatability. In this work, we show that an vibrissae-inspired vision-based magnetic-actuated whisker array integrating distributed perception with active interaction achieves reliable physical mapping, repeatable sensing, and delicate grasping, thereby advancing tactile sensing and laying the foundation for contact-driven exploration, soft manipulation, and adaptive behavior in dynamic environments.

Robot navigation in crowded environments among humans is a challenging endeavor due to its highly dynamic nature and stringent safety requirements. While existing social navigation planners typically introduce probabilistic constraints with respect to each human to provide safety guarantees, they often fail to fully exploit the collision-free space available to the robot. In this work, we explore the available collision-free space given human motion predictions and explicitly incorporate prediction uncertainty within our motion planner. To this end, we formulate motion planning as a probabilistic-inference-based optimization problem. Specifically, we perform variational inference over Gaussian Processes to generate uncertainty-aware motion plans that balance safety and efficiency. We interleave natural gradient updates with stochastic optimization to maximize the evidence lower bound, enabling efficient inference. By exploiting the sparsity inherent to dynamic models that follow the Markovian property, we achieve real-time performance and execute the generated motion plans in a receding horizon manner. Experimental results demonstrate that our approach generates smoother, efficient trajectories while maintaining safety in dynamic human environments.

Vivek Saini, Denver Haycock, William F Schneider and Alexander Dowling

Low-temperature plasma processes are increasingly employed in chemical synthesis, energy conversion, and environmental applications. Despite their promise, developing predictive models for plasma-assisted reaction systems remains challenging due to complex reaction networks, nonlinear plasma-chemistry couplings, and the high computational cost associated with detailed kinetic simulations. Traditional design strategies, such as one-factor-at-a-time studies or classical design-of-experiments (DOE) approaches, are inefficient and often neglect model uncertainty, limiting the ability to optimize reactor performance and understand system behavior.

To address these challenges, this work develops an uncertainty-aware, data-driven framework that integrates Gaussian Process (GP) regression with Bayesian Optimization (BO) to systematically explore and model plasma kinetic reaction conditions. The approach combines a physics-based plasma kinetics simulator with probabilistic surrogate modeling to reduce computational cost while preserving mechanistic fidelity. The high-fidelity ZDPlasKin kinetic model is used to simulate plasma-driven ethane dehydrogenation. Representative initial datasets are generated using Latin Hypercube Sampling of a multidimensional reactor parameter space, including reduced electric field, electron density, pressure, temperature, residence time, feed composition, and pulsed generator metrics.

These simulation data are used to train GP surrogate models that approximate nonlinear input–output relationships such as ethylene yield, selectivity, and energy efficiency while simultaneously providing predictive uncertainty estimates. Multiple kernel families (RBF, Matérn, Rational Quadratic) and hyperparameter learning strategies are evaluated to quantify their impact on model accuracy and uncertainty calibration. The flexibility of GP surrogates allows them to capture the stiff, highly nonlinear behavior characteristic of plasma reaction systems without requiring excessively large training sets.

The GP posterior uncertainty is then used to guide Bayesian Optimization, employing acquisition functions such as Expected Improvement, Upper Confidence Bound, and Probability of Improvement to balance exploration of uncertain regions and exploitation of promising operating conditions. Each BO-recommended simulation point is evaluated with ZDPlasKin and used to update the GP model, creating a closed-loop refinement process. This iterative workflow concentrates simulations in the most informative regions of the design space and significantly reduces the number of full-scale kinetic simulations required to identify near-optimal reactor conditions.

Results for maximizing ethane-to-ethylene conversion demonstrate that the GP–BO framework accurately captures key plasma kinetic trends while reducing simulation demand by more than two-thirds compared to conventional DOE and random sampling approaches. Overall, the study shows that integrating surrogate modeling, predictive uncertainty quantification, and adaptive sampling provides a practical and computationally efficient strategy for exploring high-dimensional plasma reactor conditions. The methodology offers a generalizable framework for data-efficient mechanism exploration and model-guided optimization in plasma-assisted catalysis and other complex reacting systems where direct simulation is expensive.

While sea turtle hatchlings possess flippers finely tuned for efficient aquatic propulsion, these limbs must also manage effective traversal of loose terrestrial substrates during their essential journey from nest to ocean. Hatchling's ability to generate rapid, coordinated, cyclic, and repeatable flipper-driven strokes on yielding terrain has inspired robotic systems designed to operate in similarly deformable environments. Central Pattern Generators (CPGs), modeled after the neural circuits that produce rhythmic motion in animals, provide a biologically grounded framework for coordinating these cyclic limb movements. Identifying the CPG parameters that most strongly influence locomotor performance is essential for developing adaptive and energy-efficient gait controllers. In this study, we present a detailed parameter-sensitivity analysis of a Hopf-oscillator Central Pattern Generator (CPG) that drives a sea-turtle-inspired robot in a MuJoCo simulation environment. We performed systematic sweeps of key CPG parameters, including stance and swing frequencies, amplitude gains, and inter-oscillator coupling strength, quantifying their effects on forward speed, mechanical energy expenditure, and the cost of transport. The results demonstrate that stance frequency is the dominant determinant of forward propulsion, with speed rapidly increasing up to approximately 1.0-1.1Hz before saturating. Amplitude analysis showed that foreflipper amplitude is crucial, producing over a twofold speed increase, while hindflipper and hip amplitudes primarily influence stability and energy consumption. Furthermore, the coupling strength was found to strongly modulate the quality of flipper coordination. Overall, these findings identify the key CPG parameters governing robust flipper-based locomotion on yielding terrain, providing critical insight for future optimization and adaptive control strategies in bio-inspired robotics.

Certifying the robustness of neural networks is crucial in safety-critical learning systems. A common approach involves bounding the network’s Lipschitz constant, which quantifies the worst case sensitivity to input perturbations. However, computing the exact Lipschitz constant is NP-hard, and existing estimation methods, often based on large semidefinite programs (SDPs), scale poorly with network size. This talk presents a suite of efficient and compositional algorithms for estimating Lipschitz constants in deep feedforward neural networks.

We introduce a novel decomposition framework that reformulates the SDP-based verification problem into a sequence of small, layer wise subproblems. This compositional approach enables two estimation modes: one that solves small SDPs at each stage to provide accurate bounds, and another further relaxes the subproblems to yield closed form estimates, enabling near instantaneous estimation. We first develop ECLipsE, which operates under the assumption that activation functions have a slope lower bound of zero, and provides global Lipschitz estimates through this efficient decomposition. Building on this foundation, we introduce ECLipsE-Gen-Local, a generalized framework that supports heterogeneous activation slope bounds, arbitrary input-output node pairs, and custom subnetwork configurations. It further incorporates local input information and propagates it along with the subproblem solutions to yield local Lipschitz estimates that accurately capture local features, making it especially useful in control and online learning settings.

Both algorithms come with theoretical guarantees on the feasibility and validity of Lipschitz bounds. Experimental results show that these methods outperform existing techniques across various benchmarks, offering substantial reductions in computation time (by several orders of magnitude, while maintaining or improving the tightness of the estimated bounds. In particular, the local estimation variants accurately capture region-specific features of the network's behavior, making the approach especially suited for scalable, certifiable deep learning applications where efficiency and robustness are essential.

Embodied intelligent networks are collections of distributed autonomous agents that can sense, reason, communicate, and act. Scalable and reliable coordination among such agents can benefit society in tasks ranging from environmental monitoring to transportation to surveillance. But achieving scalability is challenging due to the agents’ limited resources vs. their resource-demanding tasks, often combinatorial and NP-hard. Achieving reliability is challenging due to environmental unpredictability, limited environmental observability, and untrustworthiness of commands externally suggested by human operators or machine learning algorithms. This thesis lays the theoretical and algorithmic foundation to overcome these challenges by introducing discrete optimization and online learning capabilities that enable multi-agent networks to self-configure their communication topology to balance the trade-off of scalability vs. coordination performance, adapt online to unpredictable environments, and robustly benefit from untrustworthy external commands. The provided methods are evaluated in information-gathering tasks of mapping, target tracking, and surveillance via both physics-based simulations and field experiments.

In this conference, we would like to demonstrate a bio-inspired flight controller and robotic platform that enables bird-scale flapping wing vehicles to have longer, and more efficient flights. In nature, we often see seagulls and albatrosses gliding (without flapping) effortlessly in gusts of wind near the seashore. Birds such as albatrosses are able to fly tirelessly for days without needing to rest. The secret lies in the unique flight pattern called “Dynamic Soaring”. By utilizing the wind shear layers above the ocean surface and their aerodynamic body structure, seagulls and albatrosses are able to easily gain altitude without expending energy (through active flapping). Unconventionally, by treating the wind gusts as means to lower their total energy output instead of treating them as disturbance, albatrosses can perform energy efficient endurance flights. Inspired by dynamic soaring and its unique application in increasing flight efficiency, we will propose a new bio-inspired flapping wing vehicle platform with a flight control scheme in this conference and provide live demonstrations of this new experimental platform.

We present a novel class of minimax optimal control problems with positive dynamics, linear objective function and homogeneous constraints. The proposed problem class can be analyzed with dynamic programming and an explicit solution to the Bellman equation can be obtained, revealing that the optimal control policy (among all possible policies) is linear. This policy can in turn be computed through standard value iterations. Moreover, the feedback matrix of the optimal controller inherits the sparsity structure from the constraint matrix of the problem statement. This permits structural controller constraints in the problem design and simplifies the application to large-scale systems. We use a simple example of voltage control in an electric network to illustrate the problem setup.