Организация маршрута анимата на основе визуальных ориентиров
A mechanism for memorizing the route that a mobile robot passes in search of a target object is considered. The basis of the proposed method is to memorize the path by visual reference points and fuzzy control. The benchmark is a compact group of objects that differ in color and size. A hierarchical procedure for recognizing landmarks and scenes is described. An algorithm for forming the route description is proposed. The rules for its interpretation include elements of spatial logic. The results of simulation obtained using the Kvorum modeling system are presented.
The language RCC8RCC8is a widely-studied formalism for describing topological arrangements of spatial regions. The variables of this language range over the collection of non-empty, regular closed sets of n -dimensional Euclidean space, here denoted RC+(Rn)RC+(Rn), and its non-logical primitives allow us to specify how the interiors, exteriors and boundaries of these sets intersect. The key question is the satisfiability problem : given a finite set of atomic RCC8RCC8-constraints in m variables, determine whether there exists an m -tuple of elements of RC+(Rn)RC+(Rn)satisfying them. These problems are known to coincide for all n≥1n≥1, so that RCC8RCC8-satisfiability is independent of dimension. This common satisfiability problem is NLogSpace-complete. Unfortunately, RCC8RCC8lacks the means to say that a spatial region comprises a ‘single piece’, and the present article investigates what happens when this facility is added. We consider two extensions of RCC8RCC8: RCC8cRCC8c, in which we can state that a region is connected , and RCC8c∘RCC8c∘, in which we can instead state that a region has a connected interior. The satisfiability problems for both these languages are easily seen to depend on the dimension n , for n≤3n≤3. Furthermore, in the case of RCC8c∘RCC8c∘, we show that there exist finite sets of constraints that are satisfiable over RC+(R2)RC+(R2), but only by ‘wild’ regions having no possible physical meaning. This prompts us to consider interpretations over the more restrictive domain of non-empty, regular closed, polyhedral sets, RCP+(Rn)RCP+(Rn). We show that (a) the satisfiability problems for RCC8cRCC8c(equivalently, RCC8c∘RCC8c∘) over RC+(R)RC+(R)and RCP+(R)RCP+(R)are distinct and both NP-complete; (b) the satisfiability problems for RCC8cRCC8cover RC+(R2)RC+(R2)and RCP+(R2)RCP+(R2)are identical and NP-complete; (c) the satisfiability problems for RCC8c∘RCC8c∘over RC+(R2)RC+(R2)and RCP+(R2)RCP+(R2)are distinct, and the latter is NP-complete. Decidability of the satisfiability problem for RCC8c∘RCC8c∘over RC+(R2)RC+(R2)is open. For n≥3n≥3, RCC8cRCC8cand RCC8c∘RCC8c∘are not interestingly different from RCC8RCC8. We finish by answering the following question: given that a set of RCC8cRCC8c- or RCC8c∘RCC8c∘-constraints is satisfiable over RC+(Rn)RC+(Rn)or RCP+(Rn)RCP+(Rn), how complex is the simplest satisfying assignment? In particular, we exhibit, for both languages, a sequence of constraints ΦnΦn, satisfiable over RCP+(R2)RCP+(R2), such that the size of ΦnΦngrows polynomially in n , while the smallest configuration of polygons satisfying ΦnΦn cuts the plane into a number of pieces that grows exponentially. We further show that, over RC+(R2)RC+(R2), RCC8cRCC8c again requires exponentially large satisfying diagrams, while RCC8c∘RCC8c∘ can force regions in satisfying configurations to have infinitely many components.
We consider the quantifier-free languages, Bc and Bc°, obtained by augmenting the signature of Boolean algebras with a unary predicate representing, respectively, the property of being connected, and the property of having a connected interior. These languages are interpreted over the regular closed sets of Rn (n ≥ 2) and, additionally, over the regular closed semilinear sets of Rn. The resulting logics are examples of formalisms that have recently been proposed in the Artificial Intelligence literature under the rubric Qualitative Spatial Reasoning. We prove that the satisfiability problem for Bc is undecidable over the regular closed semilinear sets in all dimensions greater than 1, and that the satisfiability problem for Bc and Bc° is undecidable over both the regular closed sets and the regular closed semilinear sets in the Euclidean plane. However, we also prove that the satisfiability problem for Bc° is NP-complete over the regular closed sets in all dimensions greater than 2, while the corresponding problem for the regular closed semilinear sets is ExpTime-complete. Our results show, in particular, that spatial reasoning is much harder over Euclidean spaces than over arbitrary topological spaces.
In this paper we consider general scene recognition problem for analysis of user preferences based on his or her photos on mobile phone. Special attention is paid to out-of-class detections and efficient processing using MobileNet-based architectures. We propose the three stage procedure. At first, pre-trained convolutional neural network (CNN) is used extraction of input image embeddings at one of the last layers, which are used for training a classifier, e.g., support vector machine or random forest. Secondly, we fine-tune the pre-trained network on the given training set and compute the predictions (scores) at the output of the resulted CNN. Finally, we perform object detection in the input image, and the resulted sparse vector of detected objects is classified. The decision is made based on a computation of a weighted sum of the class posterior probabilities estimated by all three classifiers. Experimental results with a subset of ImageNet dataset demonstrate that the proposed approach is up to 5% more accurate when compared to conventional fine-tuned models.
The paper considers the problem of controlling a robot using a voice interface with speech recognition and analysis of the resulting set of words. The proposed method of command recognition is based on a dictionary of commands and special modifier words that are used for sentiment analysis of the command phrase and determining the priority of the task execution.
The mobile robots control subsystems are presented.