In this paper, we propose a novel Deep Reinforcement Learning (DRL) algorithm which can navigate non-holonomic robots with continuous control in an unknown dynamic environment moving obstacles. We call the approach MK-A3C (Memory and Knowledge-based Asynchronous Advantage Actor-Critic) for short. As its first component, builds GRU-based memory neural network to enhance robot's capability temporal reasoning. Robots without it tend suffer from lack of rationality face incomplete noisy...

Visual grounding is a task that seeks to predict the specific location of an object or region described by linguistic expression within image. Despite recent success, existing methods still suffer from two problems. First, most use independently pre-trained unimodal feature encoders for extracting expressive embeddings, thus resulting in significant semantic gap between embeddings and limiting effective interaction visual-linguistic contexts. Second, attention-based approaches equipped with...

Since an individual approach can hardly navigate robots through complex environments, we present a novel two-level hierarchical framework called JPS-IA3C (Jump Point Search improved Asynchronous Advantage Actor-Critic) in this paper for robot navigation dynamic environments continuous controlling signals. Its global planner JPS+ (P) is variant of JPS Search), which efficiently computes abstract path neighboring jump points. These nodes, are seen as subgoals, completely rid Deep Reinforcement...

Jump Point Search (JPS) is a well known symmetry-breaking algorithm that can substantially improve performance for grid-based optimal pathfinding. When the input grid static further speedups be obtained by combining JPS with goal bounding techniques such as Geometric Containers (instantiated Bounding Boxes) and Compressed Path Databases. Two methods, JPS+BB Two-Oracle PlannING (Topping), are currently among fastest approaches computing shortest paths on grids. The principal drawback these...

Formation control of multirobot systems has drawn significant attention in the recent years. This paper presents a potential field algorithm, navigating swarm robots into predefined 2D shape while avoiding intermember collisions. The algorithm applies both stationary and moving targets formation. We define bounded artificial forces form exponential functions, so that behavior drove by can be adjusted via selecting proper parameters. theoretical analysis proves stability convergence...

Deceptive path-planning is the task of finding a path so as to minimize probability an observer (or defender) identifying observed agent’s final goal before has been reached. It one important approaches solving real-world challenges, such public security, strategic transportation, and logistics. Existing methods either cannot make full use entire environments’ information, or lack enough flexibility for balancing path’s deceptivity available moving resource. In this work, building on recent...

In this paper, we present a hierarchical path planning framework called SG-RL (subgoal graphs-reinforcement learning), to plan rational paths for agents maneuvering in continuous and uncertain environments. By "rational", mean (1) efficient eliminate first-move lags; (2) collision-free smooth with kinematic constraints satisfied. works two-level manner. At the first level, uses geometric path-planning method, i.e., Simple Subgoal Graphs (SSG), efficiently find optimal abstract paths, also...

Recognizing the destination of a maneuvering agent is important to create intelligent AI players in Real Time Strategy (RTS) games. Among different ways problem formulation, goal recognition can be solved as model-based planning using off-the-shelf planners. However, common these frameworks that they usually lack modeling action duration real-world scenarios may take several steps transit between grids. To solve this problem, semi-Markov decision model (SMDM), which explicitly models an...

Weapon-Target Assignment (WTA) study is always a hot research topic. The model description and corresponding algorithms are keys to the success of solving WTA problems. Accordingly, paper first introduces two important concepts Weapon System Systems (WSOS) Combat Capability model. Further, based on Capabilities proposed, in which we not only consider damage enemy but also losses our own part. In addition, employ an advanced Genetic Algorithm with auto-adaptive crossover, mutations operators...

In the context of robotics and game AI, grid-based Distance Maps (DMs) are often used to fulfill collision checks by providing each traversable cell maximal clearance its closest obstacle. A key challenge for DMs’ application is how improve efficiency updating distance values when states changed (i.e., changes caused newly inserted or removed obstacles). To this end, paper presents a novel algorithm speed up construction DMs on planar, eight-connected grids. The novelty our algorithm,...

Formation control of the multi-agent system has drawn significant researches in last several decades. Concerning time consuming, local-minima trap and inflexibility team formation, this paper presents a modified potential field algorithm, transforming agents into predefined 2D-shape team. First we design bounded-exponential artificial force which consumed could be adjusted. It then carries out model obstacle avoidance, comprising vertical tangential with consideration target impacts, order...

In the context of robotics, grid-based Generalized Voronoi Diagrams (GVDs) are widely used by mobile robots to represent their surrounding area. Current approaches for incrementally constructing GVDs mainly focus on providing metric skeletons underlying grids, while connectivity among GVD vertices and edges remains implicit, which makes high-level spatial reasoning tasks impractical. this paper, we present an algorithm named Dynamic Topology Detector (DTD) extracting a with topological...

In robotics, Generalized Voronoi Diagrams (GVDs) are widely used by mobile robots to represent the spatial topologies of their surrounding area. this paper we consider problem constructing GVDs on discrete environments. Several algorithms that solve exist in literature, notably Brushfire algorithm and its improved versions which possess local repair mechanism. However, when area be processed is very large or high resolution, size metric matrices these compute can prohibitive. To address...

Global Virtual Time computation of Parallel Discrete Event Simulation is crucial for conducting fossil collection and detecting the termination simulation. The triggering condition GVT in typical approaches generally based on wall-clock time or logical intervals. However, value depends timestamps events rather than Therefore, it difficult existing to select appropriate intervals compute value. In this study, we propose a scalable estimation algorithm Lower Bound Event-Bulk-Time, which...

Intention recognition (IR) is significant for creating humanlike and intellectual agents in simulation systems. Previous widely used probabilistic graphical methods such as hidden Markov models (HMMs) cannot handle unstructural data, so logical (LHMMs) are proposed by combining HMMs first order logic. Logical semi-Markov (LHSMMs) further extend LHMMs modeling duration of states explicitly relax the assumption. In this paper, LHSMMs multi-agent intention (MAIR), which identifies not only...

Geostatistics data in regions always have highly spatial heterogeneous, yet the regional features of itself cannot be ignored. In this paper, a novel latent Bayesian model is proposed, which incorporates dependence different subjects and random effects to further analysis influence effect. The verified compatible with integrated nested Laplace approximation (INLA) framework fitted using INLA stochastic partial differential equation (SPDE). posterior marginal distribution parameters estimated...

The JPS family of grid-based pathfinding algorithms can be improved with preprocessing methods such as Geometric Containers. However, enhancements require a Dijkstra search for every node in the grid and space time costs all this additional computation prohibitive. In work we consider an alternative approach where run only from jump point is located. We also compute store geometric containers those outgoing edges which are consistent diagonal-first ordering JPS. Since number points on...

The Logical hidden Markov model (LHMM) is a combination of the first-order logic and Model (HMM). As branch statistical relational learning, LHMM great potential in many fields. In this paper, we combine logical definitions with particle filtering (PF), propose (LPF) algorithm to filter states partially missing observations. To reduce cost time, parallel resampling (LPF-PR) further proposed. experiments, an existed case about UNIX commands used test performances LPF LPF-PR. results prove...

Recognizing the destination of a moving agent is quite significant in many systems such as real time strategy games. Probabilistic graphical models are widely used to solve this problem, but existing cannot recognize changeable with noisy and partially missing observations grid based map. To two-layer semi-Markov model (TLSMM) proposed. In model, two layers represent transition destinations grids where respectively; duration being one modeled by discrete Coxian distribution. The particle...