This paper explores using ChatGPT, a large language model (LLM), as a high-level planner for multi-robot systems. We investigate whether ChatGPT can coordinate multiple service robots under spatial and operational constraints by generating movement plans and task allocations through prompt interaction alone. To this end, we design five prompt strategies, such as batch, interactive, and algorithmic, and apply them to a simulated cocktail-serving task on a 5times8 grid. Robots must serve fixed guests while avoiding collisions and following adjacency and return-home rules. ChatGPT receives only natural language prompts and is evaluated on rule compliance, coordination quality, and planning efficiency. Results show that batch prompts produce generally valid but uneven plans, while interactive prompts suffer from inconsistency and loss of shared context, especially in multi-agent settings. Algorithmic prompting, where the model generates symbolic logic, enables more structured and cooperative behavior. These findings indicate that LLMs can act as high-level reasoning agents when appropriately prompted and are best suited for generating symbolic strategies rather than executing reactive control. The study highlights the potential of LLMs as cognitive modules in hybrid or cloud-based robot systems.

This paper presents the design, implementation, and evaluation of a vision-based Detect and Avoid (DAA) system for small Uncrewed Aircraft Systems (UAS) using monocular vision and deep learning. The proposed system integrates YOLOv8 for object detection, BoT-SORT for tracking, and a rule-based avoidance strategy based on the right-hand traffic convention. The system is deployed on RyzeTech Tello EDU drones with off-board image processing and tested in indoor flight scenarios across multiple approach speeds. Experimental results demonstrate a detection accuracy of 91.7% and an 80% success rate for avoidance maneuvers in non-cooperative environments. Detection and avoidance timing varied by speed, with greater delays at lower speeds and more consistent performance at higher speeds. The system successfully tracked intruders between 6.55 and 9.25 meters, confirming its effectiveness in short-range encounters. Limitations include a narrow field of view, latency due to ground-based processing, and reduced performance in lateral approaches. Future improvements will target fisheye-compatible detection, on-board inference acceleration, and adaptive avoidance logic. These findings support the feasibility of lightweight, vision-only DAA systems for infrastructure-less, low-altitude airspace.

We propose BST-ID, a variable-length binary spatiotemporal identifier tailored for hierarchical and bandwidth-constrained systems such as LoRa-based drone networks. Unlike traditional spatial indices that treat time as an auxiliary attribute, BST-ID encodes four-dimensional coordinates (x, y, h, t) with independently adjustable resolution, enabling compact and semantically adaptive representation. The identifier includes a binary header and bit-encoded coordinates, enabling prefix filtering and low-overhead communication with anisotropic control. The format is suitable for edge devices operating in multi-hop LPWAN environments. We implemented BST-ID and evaluated it against GeoHash and Google S2, demonstrating compactness (13 bytes average), encoding speed, and flexible prefix queries. Furthermore, we discuss its potential for decentralized data exchange, including blockchain integration and Lightning Network micropayments, where each ID acts as a hash-stable, time-scoped transaction reference. BST-ID thus offers a scalable, query-efficient, and token-aware foundation for distributed sensing, routing, and autonomous information sharing.

Computation offloading for cloud robotics is receiving considerable attention in academic and industrial communities. However, current solutions face challenges: 1) traditional approaches do not consider the characteristics of networked cloud robotics (NCR) (e.g., heterogeneity and robotic cooperation); 2) they fail to capture the characteristics of tasks in a robotic streaming workflow (RSW) (e.g., strict latency requirements and varying task semantics); and 3) they do not consider quality-of-service (QoS) issues for cloud robotics. In this paper, we address these issues by proposing a QoS-aware RSW allocation algorithm for NCR with joint optimization of latency, energy efficiency, and cost, while considering the characteristics of both RSW and NCR. We first propose a novel framework that combines individual robots, robot clusters, and a remote cloud for computation offloading. We then formulate the joint QoS optimization problem for RSW allocation in NCR while considering latency, energy consumption, and operating cost, and show that the problem is NP-hard. Next, we construct a data flow graph based on the characteristics of RSW and NCR, and transform the RSW allocation problem into a mixed-integer linear programming problem. To obtain a near-optimal solution in reasonable time, we also develop a heuristic algorithm. Experiments comparing our approach with others demonstrate significant performance gains, with improved QoS and reduced execution times.

Disease similarity is a useful measure that has potential application to various aspects of medicine. One such application is the mapping of diseases in a two-dimensional plane, which can be the foundation of a useful diagnostic reminder method called the “pivot and cluster strategy.” However, the mapping of diseases using a similarity measure has yet to be explored. This article investigates such a mapping, and quantifies its basic characteristics. We first collected mutual similarity data for 1,550 diseases using a machine learning approach. The calculated similarity data were then used to map the diseases using a “multidimensional scaling” algorithm. Quantitative analysis indicated that it is difficult to express all the diseases on the map and yet still show the similarity information between the items. Then, by restricting the input, the algorithm performed well in practice. To our knowledge, this is the first study to investigate the automated mapping of diseases on a plane for use in clinical practice.

This paper describes a generalized software interface for formation flying by drones from different manufacturers. Conventional research into formation flight assumes that the drones all have the same power and functionality. However, consider a disaster response, where we might assemble a platoon of drones to sense the environment and to search for survivors by combining the different functions of drones provided by different manufacturers. The difficulties of controlling formation flight by such a variety of drones include both different mechanical specifications and different interfaces from the manufacturers for activating the same command. In this research, we construct a generalized interface for drones from each manufacturer using OpenRTM-aist. We can then assemble these drones and establish formation flight by using a virtual leader-follower system. The leader and the follower positions are calculated by using speed and rotation data from feedback information such as the GPS, velocity and rotation data from each individual machine. We also investigate good features of flight commands that can express the attributes of the representative motion of the drones. From our experiments, we show that we can establish formation flight using drones of different power and from multiple manufacturers.

We report on the recognition from a video of isolated alphabetic characters and connected cursive characters, such as Hiragana or Kanji characters, drawn in the air. This topic involves a number of difficult problems in computer vision such as the segmentation and recognition of complex motion from a video. We utilize an algorithm called time-space continuous dynamic programming (TSCDP) that can realize both time and location-free (spotting) recognition. Spotting means that prior segmentation of the input video is not required. Each of the reference (model) characters used is represented by a single stroke composed of pixels. We conducted two experiments involving the recognition of 26 isolated alphabetic characters and 23 Japanese Hiragana and Kanji air-drawn characters. Moreover we conducted gesture recognition based on TSCDP and showed that TSCDP was free from many restrictions required for conventional methods.

Web service filtering is an efficient approach to address some big challenges in service computing, such as discovery, clustering and recommendation. The key operation of the filtering process is measuring the similarity of services. Several methods are used in current similarity calculation approaches such as string-based, corpus-based, knowledge-based and hybrid methods. These approaches do not consider domain-specific contexts in measuring similarity because they have failed to capture the semantic similarity of Web services in a given domain and this has affected their filtering performance. In this paper, we propose a context-aware similarity method that uses a support vector machine and a domain dataset from a context-specific search engine query. Our filtering approach uses a spherical associated keyword space algorithm that projects filtering results from a three-dimensional sphere to a two-dimensional (2D) spherical surface for 2D visualization. Experimental results show that our filtering approach works efficiently.

This paper presents a new method for categorical classification. A method called two-dimensional continuous dynamic programming (2DCDP) is adopted to optimally capture the corresponding pixels within nonlinearly matched areas in an input image and a reference image representing an object without advance segmentation procedure. Then an image can be converted into a direction pattern which is made by matching pixels between a reference image and an input image. Finally, the category of the test image is deemed to be that which has the strongest correlation with the learning images. Experimental results show that the proposed method achieves a competitive performance on the Caltech 101 image dataset.