Large-scale online ride-sharing platforms have substantially transformed our lives by reallocating transportation resources to alleviate traffic congestion and promote efficiency. An efficient fleet management strategy not only can significantly improve the utilization of but also increase revenue customer satisfaction. It is a challenging task design an effective that adapt environment involving complex dynamics between demand supply. Existing studies usually work on simplified problem...

We present a novel order dispatch algorithm in large-scale on-demand ride-hailing platforms. While traditional approaches usually focus on immediate customer satisfaction, the proposed is designed to provide more efficient way optimize resource utilization and user experience global farsighted view. In particular, we model as sequential decision-making problem, where decision of assigning an driver determined by centralized coordinated way. The problem solved learning planning manner: 1)...

Recent works on ride-sharing order dispatching have highlighted the importance of taking into account both spatial and temporal dynamics in process for improving transportation system efficiency. At same time, deep reinforcement learning has advanced to point where it achieves superhuman performance a number fields. In this work, we propose based solution conduct large scale online A/B tests DiDi's ride-dispatching platform show that proposed method significant improvement total driver...

Order dispatching is instrumental to the marketplace engine of a large-scale ride-hailing platform, such as DiDi which continuously matches passenger trip requests drivers at scale tens millions per day. Because dynamic and stochastic nature supply demand in this context, order-dispatching problem challenging solve for an optimal solution. Added complexity are considerations system response time, reliability, multiple objectives. In paper, we describe how our approach optimization has...

Ever since the outbreak of COVID-19 epidemic, various public health control strategies have been proposed and tested against coronavirus SARS-CoV-2. We study three specific epidemic models: susceptible, exposed, infectious, recovered (SEIR) model with vaccination control; SEIR shield immunity un-quarantined infected, quarantined confirmed infected (SUQC) quarantine control. express requirement in metric temporal logic (MTL) formulas (a type formal specification languages) which can specify...

In this paper, we define a novel census signal temporal logic (CensusSTL) that focuses on the number of agents in different subsets group complete certain task specified by (STL). CensusSTL consists an "inner logic" STL formula and "outer formula. We present new inference algorithm to infer formulae from trajectory data agents. first identify then subgroups based whether agents' behaviors satisfy at each time point. use two approaches similarity complementarity, respectively. The is inferred...

Incorporating high-level knowledge is an effective way to expedite reinforcement learning (RL), especially for complex tasks with sparse rewards. We investigate RL problem where the in form of reward machines, a type Mealy machines that encode non-Markovian functions. focus on setting which this priori not available agent. develop iterative algorithm performs joint inference and policies (more specifically, q-learning). In each iteration, maintains hypothesis machine sample episodes. It uses...

Transferring high-level knowledge from a source task to target is an effective way expedite reinforcement learning (RL). For example, propositional logic and first-order have been used as representations of such knowledge. We study the transfer between tasks in which timing events matters. call temporal tasks. concretize similarity through notion logical transferability, develop approach different yet similar first propose inference technique extract metric interval (MITL) formulas...

We investigate the problem of autonomous racing among teams cooperative agents that are subject to realistic rules. Our work extends previous research on hierarchical control in head-to-head by considering a generalized version while maintaining two-level structure. A high-level tactical planner constructs discrete game encodes complex rules using simplified dynamics produce sequence target waypoints. The low-level path uses these waypoints as reference trajectory and computes...

In this paper, we present a method to learn (infer) and refine set of advices from the trajectories generated in successful failed attempts task or game, form advisory signal temporal logic (STL) formulas. Each advice consists an motion STL formula that characterizes spatial-temporal pattern as feature success selection criterion for environment select advice. For inference formulas, provide theoretical framework perfect classification with labeled different time lengths. We design...

Kaposiform hemangioendothelioma (KHE) is a rare but aggressive vascular tumor, potentially life-threatening when associated with Kasabach-Merritt phenomenon (KMP). Oral sirolimus effective may cause systemic adverse effects in infants. propranolol offers safer alternative early infancy, its efficacy plateau over time. Sequential topical enhance outcomes while minimizing toxicity. To evaluate the additive therapeutic effect and safety of KHE patients suboptimal response after oral...

We present an energy storage controller synthesis method for power systems with respect to metric temporal logic (MTL) specifications. The both constant impedance loads and are modeled as a set of differential-algebraic equations. After fault is cleared, uncertainties in the clearing time, generator machine angles rotor speed deviations will enter postfault initial states. use robust neighborhood approach cover this using neighborhoods finitely many simulated trajectories. These trajectories...

Temporal logics are widely used to express (desired) system properties in controller synthesis and verification. In linear temporal logics, the semantics of formulae defined on execution trajectories system. Recently, there have been a lot interest using dense-time logic, such as Signal Logic (STL) characterizing trajectories. this paper, we present new method derive an STL formula that characterizes motion robot arm. Our work generalizes earlier area by (i) allowing use polyhedral...

This paper investigates the problem of inferring knowledge from data that is interpretable and informative to humans who have prior knowledge. Specifically, given a dataset as collection system trajectories, we infer parametric linear temporal logic (pLTL) formulas are satisfied by trajectories in with high probability. The informativeness inferred formula measured information gain respect represented probability distribution. We first present two algorithms compute focus on types...

This paper develops a controller synthesis approach for multi-agent system (MAS) with intermittent communication. We adopt leader-follower scheme, where mobile leader absolute position sensors switches among set of followers without to provide each follower state information. model the MAS as switched system. The are asymptotically reach predetermined consensus state. To guarantee stability and followers, we derive maximum minimal dwell-time conditions constrain intervals between consecutive...

Inferring spatial-temporal properties from data is important for many complex systems, such as additive manufacturing swarm robotic systems and biological networks. Such can often be modeled a labeled graph where labels on the nodes edges represent relevant measurements temperatures distances. We introduce temporal logic (GTL) which express "whenever node's label above 10, next 3 time units there are always at least two neighboring with an edge of most 2 node 5". This paper first attempt to...

E-hailing platforms have become an important component of public transportation in recent years. The supply (online drivers) and demand (passenger requests) are intrinsically imbalanced because the pattern human behavior, especially time locations such as peak hours train stations. Hence, how to balance is one key problems satisfy passengers drivers increase social welfare. As intuitive effective approach address this problem, driver repositioning has been employed by some real-world...

We propose a method for discriminating among competing models biological systems. Our approach is based on learning temporal logic formulas from data obtained by simulating the models. apply this to find dynamic features of epidermal growth factor induced extracellular signal-regulated kinase (ERK) activation that are strictly unique positive versus negative feedback first search formula training set can eliminate ERK dynamics observed with both and then identify each model. The tested...