This letter addresses the optimal covariance control problem for stochastic discrete-time linear systems subject to chance constraints. To best of our knowledge, steering problems with probabilistic constraints have not been discussed previously in literature, although their treatment seems be a natural extension. In this letter, we first show that, unlike case no constraints, cannot decoupled mean and sub-problems. We then propose an approach solve by converting it convex programming...

This letter addresses the problem of vehicle path planning in presence obstacles and uncertainties, a fundamental robotics problem. While several algorithms have been proposed over years, many them dealt with only deterministic environments or open-loop uncertainty, i.e., uncertainty system state is not controlled and, typically, increases time because exogenous disturbances. may lead to potentially conservative nominal paths. The typical approach deal disturbances reduce use lower level...

This paper considers the problem of steering state distribution a nonlinear stochastic system from an initial Gaussian to terminal with specified mean and covariance, subject probabilistic path constraints. An algorithm is developed solve this by iteratively solving approximate linearized as convex program. method, which we call iterative covariance (iCS), numerically demonstrated controlling double integrator quadratic drag force additive Brownian noise while satisfying

Autonomous racing is a high-performance, safety-critical task that inherently involves high degree of uncertainty (especially in off-road unstructured environments), as driving conditions can vary and tire-terrain interactions are difficult to model accurately. On the one hand, vehicle needs drive fast while, on other it must avoid crashing, thus requiring tradeoff between performance safety. This work develops stochastic predictive controller (SMPC) for uncertain systems with additive...

One of the key technologies to safely operate self-driving vehicles is threat assessment other in neighborhood a vehicle. Threat algorithms must be capable predicting future movement vehicles. Many algorithms, however, predict trajectories based only on model dynamics and environment, which implies that they sometimes make too conservative predictions. This work reduces this conservativeness by capturing driver intention using random-forests classifier. Then, algorithm computes possible with...

Autonomous vehicles differ from many other autonomous systems in the sense that these have to share environment with humans (e.g., pedestrians, cyclists, and drivers traffic). This requirement poses a challenging perception planning problem. Our research focuses on problem of vehicle sharing human passenger/driver and, specifically, designing controller captures natural tendencies driver so as guarantee resulting control action is comparable human. In this paper, we propose use model into...

We address the optimal covariance steering (OCS) problem for stochastic discrete linear systems with additive Gaussian noise under state chance constraints and input hard constraints. Because system can be unbounded due to noise, are formulated as probabilistic (chance) constraints, i.e., maximum probability of constraint violation is constrained. In contrast, because it interpret appropriate control action when command violates probabilistically formulating difficult, deterministic...

We consider the problem of steering, via out-put feedback, state distribution a discrete-time, linear stochastic system from an initial Gaussian to terminal with prescribed mean and max-imum covariance, subject probabilistic path constraints on state. The filtered is obtained Kalman filter, formulated as deterministic convex program in terms observe that, presence contrast classical Linear Quadratic (LQG) control, optimal feedback control depends both process noise observation model....

This work develops a stochastic model predictive controller~(SMPC) for uncertain linear systems with additive Gaussian noise subject to state and control constraints. The proposed approach is based on the recently developed finite-horizon optimal covariance steering theory, which steers mean of system prescribed target values at given terminal time. We call our steering-based SMPC, or CS-SMPC. show that has several advantages over traditional SMPC approaches in literature. Specifically, it...

Motion-prediction algorithms for vehicles often employ historical behavior of a vehicle, rely on the Markov property underlying system, and predict future vehicle. However, alone may lead to conservative predictions heavy computational burden. To overcome these drawbacks, this paper develops method that uses notion similarity among vehicle trajectories. As traffic rules driver intentions restrict motions road is typically similar other vehicles. We hypothesize if motion any two was in past...

The purpose of this study is to apply a stochastic optimal control method the guidance an aircraft. aircraft given make final approach through microburst, hazardous weather phenomenon for low-flying that may cause accidents during landing. As determination size and strength microburst always involves uncertainty, its precise detection difficult. To minimize risk accidents, we must account system difficult task using conventional optimization techniques, which cannot handle random variables....

Many driver models assume that a human can be modeled as linear time-invariant system. Although during specific execution tasks this reasonably good model, in general, is an unrealistic and quite restrictive assumption for most real-life situations where more complex cognitive functions need to evoked, such long-term, deliberative planning, prioritization among several possible alternatives, etc. In paper we model hybrid controller switches between discrete planning short-term, continuous...

To reduce human driving workload, many advanced driver assist systems (ADAS) have been developed using a single, often simple, model to predict human-driver interaction in the immediate future. However, each person drives differently, necessitating personalized models based on data obtained from actual actions. Yet, traditional control-theoretic and physics-based difficulty accurately predicting Being inspired recent achievements of machine-learning (ML) methods, this work compares several...

This paper discusses a new guidance strategy to extend mission time of small unmanned aerial vehicles (UAVs) by harvesting energy from the atmosphere. Small UAVs are getting common for people investigate areas that too dangerous or expensive manned air execute. However, their sizes restrict payloads, which directly leads poor range and time. One solutions this problem is extract wind fields reduce fuel required fly back forth between area ground station. The speed principle optimization...

We address the optimal covariance steering (OCS) problem for stochastic discrete linear systems with additive Gaussian noise under state chance constraints and input hard constraints. Because system can be unbounded due to noise, are formulated as probabilistic (chance) constraints, i.e., maximum probability of constraint violation is constrained. In contrast, because it interpret appropriate control action when command violates probabilistically formulating difficult, deterministic...

This work proposes a new self-driving framework that uses human driver control model, whose feature-input values are extracted from images using deep convolutional neural networks (CNNs). The development of image processing techniques CNNs along with accelerated computing hardware has recently enabled real-time detection these values. use models can lead to more "natural" driving behavior vehicles. Specifically, we the well-known two-point visual model as controller, and top-down lane cost...

This paper addresses the problem of steering a discrete-time linear dynamical system from an initial Gaussian distribution to final in game-theoretic setting. One two players strives minimize quadratic payoff, while at same time tries meet given mean and covariance constraints time-step. The other player maximizes but it is assumed be indifferent terminal constraint. At first, unconstrained version game examined, necessary conditions for existence saddle point are obtained. We show that...

The ballast water exchange at seas has been recognized as one of the operational countermeasures to cope with invasion non-indigenous species through water. Bay Bengal is traditionally considered have low chlorophyll-a concentration thus phytoplankton counts, which reason why (BoB) selected a suitable exchangeable sea. However an anomalously high K(490) area was found off coast Sri Lanka during northeast monsoon in 2005, corresponds higher plankton cell densities than criterion set by...

This work addresses the optimal covariance control problem for stochastic discrete-time linear time-varying systems subject to chance constraints. Covariance steering is a steer system state Gaussian distribution another while minimizing cost function. To best of our knowledge, problems have never been discussed with probabilistic constraints although it natural extension. In this work, first we show that, unlike case no constraints, cannot decouple mean and sub-problems. Then propose an...

This work addresses the problem of vehicle path planning in presence obstacles and uncertainties, which is a fundamental robotics. While many algorithms have been proposed for decades, them dealt with only deterministic environments or open-loop uncertainty, i.e., uncertainty system state not controlled and, typically, increases time due to exogenous disturbances, leads design potentially conservative nominal paths. In order deal disturbances reduce generally, lower-level feedback controller...