A novel approach to reduced-order modeling of high-dimensional time varying systems is proposed. It leverages the formalism of the Dynamic Mode Decomposition technique together with the concept of balanced realization. It is assumed that the only information available on the system comes from input, state, and output trajectories generated by numerical simulations or recorded and estimated during experiments, thus the approach is fully data-driven. The goal is to obtain an input-output low dimensional linear model which approximates the system across its operating range. Since the dynamics of aeroservoelastic systems markedly changes in operation (e.g. due to change in flight speed or altitude), time-varying features are retained in the constructed models. This is achieved by generating a Linear Parameter-Varying representation made of a collection of state-consistent linear time-invariant reduced-order models. The algorithm formulation hinges on the idea of replacing the orthogonal projection onto the Proper Orthogonal Decomposition modes, used in Dynamic Mode Decomposition-based approaches, with a balancing oblique projection constructed entirely from data. As a consequence, the input-output information captured in the lower-dimensional representation is increased compared to other projections onto subspaces of same or lower size. Moreover, a parameter-varying projection is possible while also achieving state-consistency. The validity of the proposed approach is demonstrated on a morphing wing for airborne wind energy applications by comparing the performance against two algorithms recently proposed in the literature. Comparisons cover both prediction accuracy and performance in model predictive control applications.
We consider the effect of using approximate system predictions in event-triggered control schemes. Such approximations may result from using numerical transcription methods for solving continuous-time optimal control problems. Mesh refinement can guarantee upper bounds on the error in the differential equations which model the system dynamics. With the accuracy guarantees of a mesh refinement scheme, we show that the proposed event-triggering scheme -- which compares the measured system with approximate state predictions -- can be used with a guaranteed strictly positive inter-update time. We show that if we have knowledge of the employed transcription scheme or the approximation errors, then we can obtain better online estimates of inter-update times. We additionally detail a method of tightening constraints on the approximate system trajectory used in the nonlinear programming problem to guarantee constraint satisfaction of the continuous-time system. This is the first work to incorporate prediction accuracy in triggering metrics. Using the solution accuracy we can guarantee reliable lower bounds for inter-update times and perform solution dependent constraint tightening.
Large-scale ride-sharing systems combine real-time dispatching and routing optimization over a rolling time horizon with a model predictive control (MPC) component that relocates idle vehicles to anticipate the demand. The MPC optimization operates over a longer time horizon to compensate for the inherent myopic nature of the real-time dispatching. These longer time horizons are beneficial for the quality of relocation decisions but increase computational complexity. Consequently, the ride-sharing operators are often forced to use a relatively short time horizon. To address this computational challenge, this paper proposes a hybrid approach that combines machine learning and optimization. The machine-learning component learns the optimal solution to the MPC on the aggregated level to overcome the sparsity and high-dimensionality of the solution. The optimization component transforms the machine-learning prediction back to the original granularity through a tractable transportation model. As a consequence, the original NP-hard MPC problem is reduced to a polynomial time prediction and optimization, which allows the ride-sharing operators to consider a longer time horizon. Experimental results show that the hybrid approach achieves significantly better service quality than the MPC optimization in terms of average rider waiting time, due to its ability to model a longer horizon.
Achieving an evenly distributed fertilization spread pattern is a complex technical task. A corresponding control algorithm must account for the tractor movement, the settings of the spreader, the prescribed dosage as well as machine constraints. It dictates, in particular, the fertilization process needs be estimated ahead to achieve an optimized spread pattern. The presented work is concerned with the development of a predictive control scheme for optimized fertilizer application using modeling of the tractor moving on the field and the spread pattern in form of a crescent behind the tractor. In particular, the form of the spread pattern is modeled via four normal distributions, two for each side of the pattern. The control goal is to achieve a desired fertilizer distribution on the field. The study presents three algorithms for comparison: a one-step optimization and two approaches using model-predictive control, one with a simplified model of the spread pattern in the prediction horizon, and one with a comprehensive one model, respectively. The best results are obtained with model-predictive control using the comprehensive model.
Safety is one of the fundamental problems in robotics. Recently, one-step or multi-step optimal control problems for discrete-time nonlinear dynamical system were formulated to offer tracking stability using control Lyapunov functions (CLFs) while subject to input constraints as well as safety-critical constraints using control barrier functions (CBFs). The limitations of these existing approaches are mainly about feasibility and safety. In the existing approaches, the feasibility of the optimization and the system safety cannot be enhanced at the same time theoretically. In this paper, we propose two formulations that unifies CLFs and CBFs under the framework of nonlinear model predictive control (NMPC). In the proposed formulations, safety criteria is commonly formulated as CBF constraints and stability performance is ensured with either a terminal cost function or CLF constraints. Slack variables with relaxing technique are introduced on the CBF constraints to resolve the tradeoff between feasibility and safety so that they can be enhanced at the same. The advantages about feasibility and safety of proposed formulations compared with existing methods are analyzed theoretically and validated with numerical results.
After disasters, distribution networks have to be restored by repair, reconfiguration, and power dispatch. During the restoration process, changes can occur in real time that deviate from the situations considered in pre-designed planning strategies. That may result in the pre-designed plan to become far from optimal or even unimplementable. This paper proposes a centralized-distributed bi-level optimization method to solve the real-time restoration planning problem. The first level determines integer variables related to routing of the crews and the status of the switches using a genetic algorithm (GA), while the second level determines the dispatch of active/reactive power by using distributed model predictive control (DMPC). A novel Aitken- DMPC solver is proposed to accelerate convergence and to make the method suitable for real-time decision making. A case study based on the IEEE 123-bus system is considered, and the acceleration performance of the proposed Aitken-DMPC solver is evaluated and compared with the standard DMPC method.
Vehicle trajectory optimization is essential to ensure vehicles travel efficiently and safely. This paper presents an infrastructure assisted constrained connected automated vehicles (CAVs) trajectory optimization method on curved roads. This paper systematically formulates the problem based on a curvilinear coordinate which is flexible to model complex road geometries. Further, to deal with the spatial varying road obstacles, traffic regulations, and geometric characteristics, two-dimensional vehicle kinematics is given in a spatial formulation with exact road information provided by the infrastructure. Consequently, we applied a multi-objective model predictive control (MPC) approach to optimize the trajectories in a rolling horizon while satisfying the collision avoidances and vehicle kinematics constraints. To verify the efficiency of our method, a numerical simulation is conducted. As the results suggest, the proposed method can provide smooth vehicular trajectories, avoid road obstacles, and simultaneously follow traffic regulations, which is robust to road geometries and disturbances.
This article proposes a hierarchical learning architecture for safe data-driven control in unknown environments. We consider a constrained nonlinear dynamical system and assume the availability of state-input trajectories solving control tasks in different environments. In addition to task-invariant system state and input constraints, a parameterized environment model generates task-specific state constraints, which are satisfied by the stored trajectories. Our goal is to use these trajectories to find a safe and high-performing policy for a new task in a new, unknown environment. We propose using the stored data to learn generalizable control strategies. At each time step, based on a local forecast of the new task environment, the learned strategy consists of a target region in the state space and input constraints to guide the system evolution to the target region. These target regions are used as terminal sets by a low-level model predictive controller. We show how to i) design the target sets from past data and then ii) incorporate them into a model predictive control scheme with shifting horizon that ensures safety of the closed-loop system when performing the new task. We prove the feasibility of the resulting control policy, and apply the proposed method to robotic path planning, racing, and computer game applications.
Model-based reinforcement learning (MBRL) aims to learn model(s) of the environment dynamics that can predict the outcome of its actions. Forward application of the model yields so called imagined trajectories (sequences of action, predicted state-reward) used to optimize the set of candidate actions that maximize expected reward. The outcome, an ideal imagined trajectory or plan, is imperfect and typically MBRL relies on model predictive control (MPC) to overcome this by continuously re-planning from scratch, incurring thus major computational cost and increasing complexity in tasks with longer receding horizon. We propose uncertainty estimation methods for online evaluation of imagined trajectories to assess whether further planned actions can be trusted to deliver acceptable reward. These methods include comparing the error after performing the last action with the standard expected error and using model uncertainty to assess the deviation from expected outcomes. Additionally, we introduce methods that exploit the forward propagation of the dynamics model to evaluate if the remainder of the plan aligns with expected results and assess the remainder of the plan in terms of the expected reward. Our experiments demonstrate the effectiveness of the proposed uncertainty estimation methods by applying them to avoid unnecessary trajectory replanning in a shooting MBRL setting. Results highlight significant reduction on computational costs without sacrificing performance.
We present a framework to design nonlinear robust output feedback model predictive control (MPC) schemes that ensure constraint satisfaction under noisy output measurements and disturbances. We provide novel estimation methods to bound the magnitude of the estimation error based on: stability properties of the observer; detectability; set-membership estimation; moving horizon estimation (MHE). Robust constraint satisfaction is guaranteed by suitably incorporating these online validated bounds on the estimation error in a homothetic tube based MPC formulation. In addition, we show how the performance can be further improved by combining MHE and MPC in a single optimization problem. The framework is applicable to a general class of detectable and (incrementally) stabilizable nonlinear systems. While standard output feedback MPC schemes use offline computed worst-case bounds on the estimation error, the proposed framework utilizes online validated bounds, thus reducing conservatism and improving performance. We demonstrate the reduced conservatism of the proposed framework using a nonlinear 10-state quadrotor example.
This work considers artificial feed-forward neural networks as parametric approximators in optimal control of discrete-time systems. Two different approaches are introduced to take polytopic input constraints into account. The first approach determines (sub-)optimal inputs by the application of gradient methods. Closed-form expressions for the gradient of general neural networks with respect to their inputs are derived. The approach allows to consider state-dependent input constraints, as well as to ensure the satisfaction of state constraints by exploiting recursive reachable set computations. The second approach makes use of neural networks with softmax output units to map states into parameters, which determine (sub-)optimal inputs by a convex combination of the vertices of the input constraint set. The application of both approaches in model predictive control is discussed, and results obtained for a numerical example are used for illustration.
Economic model predictive control (EMPC) is a promising methodology for optimal operation of dynamical processes that has been shown to improve process economics considerably. However, EMPC performance relies heavily on the accuracy of the process model used. As an alternative to model-based control strategies, reinforcement learning (RL) has been investigated as a model-free control methodology, but issues regarding its safety and stability remain an open research challenge. This work presents a novel framework for integrating EMPC and RL for online model parameter estimation of a class of nonlinear systems. In this framework, EMPC optimally operates the closed loop system while maintaining closed loop stability and recursive feasibility. At the same time, to optimize the process, the RL agent continuously compares the measured state of the process with the model's predictions (nominal states), and modifies model parameters accordingly. The major advantage of this framework is its simplicity; state-of-the-art RL algorithms and EMPC schemes can be employed with minimal modifications. The performance of the proposed framework is illustrated on a network of reactions with challenging dynamics and practical significance. This framework allows control, optimization, and model correction to be performed online and continuously, making autonomous reactor operation more attainable.
Long prediction horizons in Model Predictive Control (MPC) often prove to be efficient, however, this comes with increased computational cost. Recently, a Robust Model Predictive Control (RMPC) method has been proposed which exploits models of different granularity. The prediction over the control horizon is split into short-term predictions with a detailed model using MPC and long-term predictions with a coarse model using RMPC. In many applications robustness is required for the short-term future, but in the long-term future, subject to major uncertainty and potential modeling difficulties, robust planning can lead to highly conservative solutions. We therefore propose combining RMPC on a detailed model for short-term predictions and Stochastic MPC (SMPC), with chance constraints, on a simplified model for long-term predictions. This yields decreased computational effort due to a simple model for long-term predictions, and less conservative solutions, as robustness is only required for short-term predictions. The effectiveness of the method is shown in a mobile robot collision avoidance simulation.
Manipulating deformable objects, such as ropes and clothing, is a long-standing challenge in robotics, because of their large degrees of freedom, complex non-linear dynamics, and self-occlusion in visual perception. The key difficulty is a suitable representation, rich enough to capture the object shape, dynamics for manipulation and yet simple enough to be estimated reliably from visual observations. This work aims to learn latent Graph dynamics for DefOrmable Object Manipulation (G-DOOM). G-DOOM approximates a deformable object as a sparse set of interacting keypoints, which are extracted automatically from images via unsupervised learning. It learns a graph neural network that captures abstractly the geometry and the interaction dynamics of the keypoints. To handle object self-occlusion, G-DOOM uses a recurrent neural network to track the keypoints over time and condition their interactions on the history. We then train the resulting recurrent graph dynamics model through contrastive learning in a high-fidelity simulator. For manipulation planning, G-DOOM reasons explicitly about the learned dynamics model through model-predictive control applied at each keypoint. Preliminary experiments of G-DOOM on a set of challenging rope and cloth manipulation tasks indicate strong performance, compared with state-of-the-art methods. Although trained in a simulator, G-DOOM transfers directly to a real robot for both rope and cloth manipulation.
Reinforcement Learning (RL) controllers have generated excitement within the control community. The primary advantage of RL controllers relative to existing methods is their ability to optimize uncertain systems independently of explicit assumption of process uncertainty. Recent focus on engineering applications has been directed towards the development of safe RL controllers. Previous works have proposed approaches to account for constraint satisfaction through constraint tightening from the domain of stochastic model predictive control. Here, we extend these approaches to account for plant-model mismatch. Specifically, we propose a data-driven approach that utilizes Gaussian processes for the offline simulation model and use the associated posterior uncertainty prediction to account for joint chance constraints and plant-model mismatch. The method is benchmarked against nonlinear model predictive control via case studies. The results demonstrate the ability of the methodology to account for process uncertainty, enabling satisfaction of joint chance constraints even in the presence of plant-model mismatch.
Many robotics domains use some form of nonconvex model predictive control (MPC) for planning, which sets a reduced time horizon, performs trajectory optimization, and replans at every step. The actual task typically requires a much longer horizon than is computationally tractable, and is specified via a cost function that cumulates over that full horizon. For instance, an autonomous car may have a cost function that makes a desired trade-off between efficiency, safety, and obeying traffic laws. In this work, we challenge the common assumption that the cost we optimize using MPC should be the same as the ground truth cost for the task (plus a terminal cost). MPC solvers can suffer from short planning horizons, local optima, incorrect dynamics models, and, importantly, fail to account for future replanning ability. Thus, we propose that in many tasks it could be beneficial to purposefully choose a different cost function for MPC to optimize: one that results in the MPC rollout having low ground truth cost, rather than the MPC planned trajectory. We formalize this as an optimal cost design problem, and propose a zeroth-order optimization-based approach that enables us to design optimal costs for an MPC planning robot in continuous MDPs. We test our approach in an autonomous driving domain where we find costs different from the ground truth that implicitly compensate for replanning, short horizon, incorrect dynamics models, and local minima issues. As an example, the learned cost incentivizes MPC to delay its decision until later, implicitly accounting for the fact that it will get more information in the future and be able to make a better decision. Code and videos available at https://sites.google.com/berkeley.edu/ocd-mpc/.
Economic Model Predictive Control (MPC) dissipativity theory is central to discussing the stability of policies resulting from minimizing economic stage costs. In its current form, the dissipativity theory for economic MPC applies to problems based on deterministic dynamics or to very specific classes of stochastic problems, and does not readily extend to generic Markov Decision Processes. In this paper, we clarify the core reason for this difficulty, and propose a generalization of the economic MPC dissipativity theory that circumvents it. This generalization focuses on undiscounted infinite-horizon problems and is based on nonlinear stage cost functionals, allowing one to discuss the Lyapunov asymptotic stability of policies for Markov Decision Processes in terms of the probability measures underlying their stochastic dynamics. This theory is illustrated for the stochastic Linear Quadratic Regulator with Gaussian process noise, for which a storage functional can be provided explicitly. For the sake of brevity, we limit our discussion to undiscounted Markov Decision Processes.
This paper presents two stochastic model predictive control methods for linear time-invariant systems subject to unbounded additive uncertainties. The new methods are developed by formulating the chance constraints into deterministic form, which are treated in analogy with robust constraints, by using the probabilistic reachable set. Firstly, the probabilistically resolvable time-varying tube-based stochastic model predictive control algorithm is designed by employing the time-varying probabilistic reachable sets as the tubes. Secondly, by utilizing the probabilistic positively invariant set, the probabilistically resolvable constant tube-based stochastic model predictive control algorithm is developed by employing the constantly tightened constraints in the entire prediction horizons. In addition, to enhance the feasibility of the algorithms, the soft constraints are imposed to the state initializations. The algorithm feasibility and closed-loop stability results are provided. The efficacy of the approaches are demonstrated by means of numerical simulations.
Model-free reinforcement learning (RL) for legged locomotion commonly relies on a physics simulator that can accurately predict the behaviors of every degree of freedom of the robot. In contrast, approximate reduced-order models are commonly used for many model predictive control strategies. In this work we abandon the conventional use of high-fidelity dynamics models in RL and we instead seek to understand what can be achieved when using RL with a much simpler centroidal model when applied to quadrupedal locomotion. We show that RL-based control of the accelerations of a centroidal model is surprisingly effective, when combined with a quadratic program to realize the commanded actions via ground contact forces. It allows for a simple reward structure, reduced computational costs, and robust sim-to-real transfer. We show the generality of the method by demonstrating flat-terrain gaits, stepping-stone locomotion, two-legged in-place balance, balance beam locomotion, and direct sim-to-real transfer.
This paper addresses the trajectory tracking problem of an autonomous tractor-trailer system by using a fast distributed nonlinear model predictive control algorithm in combination with nonlinear moving horizon estimation for the state and parameter estimation in which constraints on the inputs and the states can be incorporated. The proposed control algorithm is capable of driving the tractor-trailer system to any desired trajectory ensuring high control accuracy and robustness against environmental disturbances.