This work proposes a new natural language processing (NLP) task to tackle the issue of multi-round, sequential text-based knowledge update. The study introduces a hybrid learning architecture and a novel self-supervised training strategy to enable generative language models to consolidate knowledge in the same way as humans. A dataset was also created for evaluation and results showed the effectiveness of our methodology. Experimental results confirm the superiority of the proposed approach over existing models and large language models (LLMs). The proposed task and model framework have the potential to significantly improve the automation of knowledge organization, making text-based knowledge an increasingly crucial resource for powerful LLMs to perform various tasks for humans.

Keeping knowledge facts up-to-date is labored and costly as the world rapidly changes and new information emerges every second. In this work, we introduce a novel task, news event triggered knowledge update. Given an existing article about a topic with a news event about the topic, the aim of our task is to generate an updated article according to the information from the news event. We create a multi-grained dataset for the investigation of our task. The articles from Wikipedia are collected and aligned with news events at multiple language units, including the citation text, the first paragraph, and the full content of the news article. Baseline models are also explored at three levels of knowledge update, including the first paragraph, the summary, and the full content of the knowledge facts.

Creatives in animation and film productions have forever been exploring the use of new means to prototype their visual sequences before realizing them, by relying on hand-drawn storyboards, physical mockups or more recently 3D modelling and animation tools. However these 3D tools are designed in mind for dedicated animators rather than creatives such as film directors or directors of photography and remain complex to control and master. In this paper we propose a VR authoring system which provides intuitive ways of crafting visual sequences, both for expert animators and expert creatives in the animation and film industry. The proposed system is designed to reflect the traditional process through (i) a storyboarding mode that enables rapid creation of annotated still images, (ii) a previsualisation mode that enables the animation of the characters, objects and cameras, and (iii) a technical mode that enables the placement and animation of complex camera rigs (such as cameras cranes) and light rigs. Our methodology strongly relies on the benefits of VR manipulations to re-think how content creation can be performed in this specific context, typically how to animate contents in space and time. As a result, the proposed system is complimentary to existing tools, and provides a seamless back-and-forth process between all stages of previsualisation. We evaluated the tool with professional users to gather experts' perspectives on the specific benefits of VR in 3D content creation.

As more schools incorporate 3D printing into their curriculum to stimulate the creativity of K-12 students with a learning-by-doing approach, it becomes crucial to understand how users work with 3D modeling tools. In this paper, we aim to develop model and usage-pattern-related features to quantize students' performance on 3D modeling operation. The dataset is gathered from the Affiliated Experimental Elementary School (AEES) of National Chengchi University. Participants' operation log and finished work for specific 3D modeling software are recorded and analyzed. In all our lesson plans, students are required to create structurally stable and printable 3D models. Three modeling software with different levels of difficulty have been introduced and tested. The collected data include screen recording, software operation log, experts evaluation, and interviews with students, which are employed for subsequent qualitative evaluation as well as quantitative analysis. With our proposed approach, we are able to identify the key factors affecting students' learning experience and performance in terms of model completeness and usage pattern. Through these indicators, instructors can understand student's learning status of 3D modeling software more comprehensively.

In the recent years, due to the prevalence of 3D printers, there have been many applications of 3D printing developed for education such as creating custom-made teaching aids or student art work. Although there are web sites hosting 3D models created by students and allowing them to share and search in the database, these systems usually lack functions of collaborative learning. On the other hand, most learning management systems do not provide functions needed for sharing and viewing 3D models. In this work, we aim to design a system called 3D Model Co-learning Space (3D MCLS), dedicated to collaborative learning allowing students to share 3D models as well as performing blind peer assessment. The system allows a user to store, share, display, and discuss 3D models and allows a teacher to manage a group of students in a flexible way. More specifically, the system can automatically create a thumbnail by computing an optimized view of an uploaded 3D model. It uses tags to organize models into groups according to their attributes or teams in a class for easier access through search. In addition, it can provide blind assignments of peer review and compare the results with self-assessment. We have implemented such a system and conducted a pilot study to obtain a preliminary evaluation on the usability of the system.

ABSTRACT Navigation is a critical task for agents populating virtual worlds. In the last years, numerous solutions have been proposed to solve the path planning problem in order to enhance the autonomy of virtual agents. Those solutions mainly focused on static environments, eventually populated with dynamic obstacles. However, dynamic objects are usually more than just obstacles as they can be used by an agent to reach new locations. In this paper, we propose an online path planning algorithm in dynamically changing environments with unknown evolution such as physically based‐environments. Our method represents objects in terms of obstacles but also in terms of navigable surfaces. This representation allows our algorithm to find temporal paths through disconnected and moving platforms. We will also show that the proposed method also enables several kinds of adaptations such as avoiding moving obstacles or adapting the agent postures to environmental constraints. Copyright © 2012 John Wiley & Sons, Ltd.

The goal of our research is to develop a computer system based on the concept of pictures-and-attributed-notes (PAN) aiming to stimulate creativity and imagination when users create a story. We have conducted an experiment and used the Conceptual, Operational, Perceptional and Evaluation (COPE) coding system to analyze the process of creating a story with the aid of computer system. The preliminary results revealed that the computer system based on PAN can stimulate user's creativity during the process of story creation.

Automatic motion generation for digital character under the real-time user control is a challenging problem for computer graphic research and virtual environment applications such as on-line games. The present study introduces a methodology to generate a glove puppet animation which is controlled by a new input device, called the Smart Glove, capturing the hand motion. An animation system is proposed to generate the puppet animation based on the procedural animation and motion capture data from Smart Glove. As the control of the puppet character in the animation takes into account the design of Smart Glove and the operation of the puppet in reality, the physical hand motion can either activate the designed procedural animation through motion recognition or tune the parameters of the procedural animation to build the new puppet motion, which allows the direct user control on the animation. The potential application and improvement of the current animation system are also discussed.

One of the most common effects in the animated feature film Quantum Quest are the trails that fall behind and twisted around all photon characters, including the hero character Dave (Figure 1). The quantity of shots that required this effect warranted the development

Creativity plays an important role in many learning activities but how to enhance creativity with the support of modern computer tools deserves further studies. In this paper, we propose a game-based collaborative creativity support system called Idea storming cube to support creative thinking and help a user form a perspective-shift thinking habit. The system analyzes the knowledge acquired from the history of user inputs and compares it with the ideas possessed by the domain expert and other users in the current brainstorming group. The system is designed to provide user-, goal- and context-sensitive supports with this mechanism that stimulates more divergent thinking. We have implemented the proposed game with a magic-cube-like HCI for collaborative idea generation. A case study for preliminary evaluation of the system is also reported in this paper.

Interactivity is a critical issue in designing a good game or similar virtual environments. As the computer hardware is continuously improved, more computing power can be invested in interactivity in addition to graphics rendering. For games with virtual characters, how a character interact with a user is usually specified at design time according to a given scene. Consequently, the characters in a game usually can only display canned motions at designated locations of a scene. After several runs of practice, the user may easily get bored because these actions become predictable. Therefore, it is highly desirable to have a smarter character that can plan its motions according to the inputs from the user as well as other constraints from the environments.

Collision detection is a critical module in many applications such as computer graphics, robot motion planning, physical simulation, CAD/CAM, and molecular modeling. Many efficient algorithms have been proposed to solve the collision-detection problem, and most of them use some sort of hierarchical bounding volume to speed up the time-consuming process. In this research, we focus on a special situation for applications such as interactive virtual environment, where accuracy may not be crucial but the available time for performing the check is limited. We describe our findings on the inherencies of bounding volume traverse trees (BVTT). Based on this observation, we propose a time-budgeted collision detection method that can traverse the critical regions of BVTT earlier if the time budget is limited. Preliminary experimental results are also reported, and a new strategy is suggested based on the concept of confidence value

Interactive avatar control means a control mechanism using AI techniques to improve user navigation efficiency as well as visual realism in a 3D virtual environment. This paper reports a novel approach to enhancing intelligent navigation interface with a third-person control mechanism. We use the concept of CT-space roadmap to design a new motion planning algorithm that can generate collision-free motions for the upper-body of an avatar walking in a cluttered virtual environment. We have designed a mechanism to allow motion generation and execution to occur in parallel such that collision-free motions can be generated in real time. We believe that this type of intelligent interface can enhance the efficiency of avatar navigation as well as improve the realism of the generated avatar motions. The same techniques can also be applied to the motion control of humanoid robots to avoid obstacles in real time

It has been a major objective for researchers to develop computer systems that can effectively deliver instruction to learners. Therefore, how to incorporate instructional strategies in computer-assisted learning systems in a systematic manner deserves further investigation. In this paper, a style-matching strategy that attempts to match learning materials' styles to learners' latent traits is proposed and realized in a web-based tutoring system, called CooTutor. The mechanism of adaptive material selection takes learners' different spatial ability and learning styles as an integral learning profile into account, and performs traits-based personalization of learning experience. This system is specifically designed to conquer the difficulty of tutoring the topic on fundamental spatial geometry in conventional curriculums. By conducting empirical evaluation with a small group of students, it is found that CooTutor is generally beneficial to learning the domain, but the effect of the styles-matching mechanism remains inconclusive. The work aims to contribute to the community of adaptive hypermedia in providing an explorative example adopting the concern of individual difference for personalization. The system design, a usage scenario, and an exploratory evaluation are presented in this paper as implications for further studies.

Abstract The creation of a stylistic animation through the use of high‐level controls has always been a design goal for computer animation software. In this paper, we propose a procedural animation system, called rhythmic character animation playacting (RhyCAP), which allows a designer to interactively direct animated characters by adjusting rhythmic parameters such as tempo, exaggeration, and timing. The motions thus generated reflect the intention of the director and also adapt to environmental obstacle constraints. We use a sequence of martial‐art steps in the performance of a Chinese lion dance to illustrate the effectiveness of the system. The animation is generated by composition of common motion elements, concisely represented in an action graph. We have implemented an animation control program that allows Chinese lion dance to be choreographed interactively. This authoring tool also serves as a useful means for preserving this part of world cultural heritage. Copyright © 2006 John Wiley & Sons, Ltd.

Currently most systems capable of performing intelligent camera control use cinematographic idioms or a constraint satisfaction mechanism to determine a sequence of camera configurations for a given animation script. However, an automated cinematography system cannot be made practical without taking idiosyncrasy and the distinct role of each member in a filmmaking team into account. In this paper, we propose an interactive virtual cinematographer model imitating the key functions of a real filmmaking team consisting of three modules: director, photographer, and editor. The system uses parameterized cinematographic idioms in the three modules to determine the best camera configurations for an animation script. The system allows a user to interact with the virtual cinematographer to specify stylistic preferences, which can be carried over to other animation scripts.

In this paper we focus on incorporating adaptive presentation and 3D visualization into Web-based learning environments to enhance learners' learning outcomes, especially to facilitate learners' spatial reasoning on geometric topics in computer graphics. This system is called CooTutor (coordinate tutor). According to educational media studies, "Media and Method" are the main concerns of developing such a system. In CooTutor, interactive 3D media is used to present spatial relations effectively to the learners. Instructional methods and strategies are embedded in its adaptive mechanism by incorporating ITS (intelligent tutoring systems) techniques. Since geometric transformation is the example domain of our study, learners' spatial ability is considered as an index for adapting the presentation. To achieve better abstraction and flexibility, we have chosen to separate the concept sequencing from the underlying learning materials. We believe that Web-based learning could take more advantages of available computing power to enhance learning by realizing innovative instructional design, as in the case of 3D interactive presentation.

Character animations on most virtual environment systems are canned motions created off-line through motion capture techniques. The motions are then encoded and transmitted with a fixed format and played at the client side. In this paper, we have proposed an XML-based scripting language, called eXtensible Animation Markup Language (XAML), to describe interactive dialog-based animations. The language is designed to describe character animations at various command levels and to compose a new animation from existing animation clips. In addition, the language is extended to incorporate other dialog-based scripting language such as VoiceXML. We have implemented such a system in Java that can interpret the language and render 3D animations based on the user's interactive voice commands.

Enabling a humanoid robot to move autonomously in a real-life environment presents a challenging problem. Unlike traditional wheeled robots, legged robots such as humanoid robots have advanced abilities of stepping over an object or striding over a deep gap with versatile locomotions. However, only few humanoid researches today have addressed this problem. In this paper, we consider the problem of planning a humanoid robot's motion in a layered environment cluttered with obstacles and deep narrow gaps. We extend the motion planning system for humanoids in our previous work to account for multiple locomotions and striding ability for a humanoid robot. Each locomotion corresponds to a tier in the search space, and the tiers are connected at the locations where motion transitions are possible. According to a humanoid's geometric properties, such as maximal gait size and step height, we propose to apply the closing morphological operator to the workspace bitmap to compute the reachability region for a humanoid with striding ability. Our experiments show that our system is efficient in generating versatile motions for a humanoid to reach its goal in a complex environment.

The WALK mode is one of the most common navigation interfaces for 3D virtual environments. However, due to the limited view angle and low frame rate, users are often blocked by obstacles when they navigate in a cluttered virtual scene with such a mode. Intelligent 3D navigation interfaces with assisting mechanisms, such as motion planning methods or virtual force fields, have been proposed in the literature to improve navigation efficiency. Nevertheless, the applicability of these methods is subject to individual discrepancy, and the control parameters of these methods are usually determined by empirical means. In this paper, we propose an intelligent navigation interface with a personalizable assisting mechanism. We have designed two methods, simulation experiment and dynamic adjustment, to find the best control parameters for composing artificial forces for an individual in an off-line and on-line manner, respectively. The simulation experiment method searches for the optimal control parameters for a user in a systematic manner while the dynamic adjustment method makes the assisting mechanism adaptive to user control behaviors as well as environmental variations in real time. Our experiments show that both methods can further improve the navigation efficiency for a wider range of users.

Moving a crowd of robots or avatars from their current configurations to some destination area without causing collisions is a challenging motion-planning problem because the high degrees of freedom involved. Two approaches are often used for this type of problems: decoupled and centralized. The tradeoff of these two approaches is that the decoupled approach is considered faster while the centralized approach has the advantage of being complete. In this paper, we propose an efficient centralized planner that is much faster than the traditional randomized planning approaches. This planner uses a hierarchical sphere tree structure to group robots dynamically. By taking advantage of the problem characteristics on independently moving robots, we are able to design a practical planner with the centralized approach when the number of robots is rather large. We use several simulation examples to demonstrate the efficiency and effectiveness of the planner.

Motion planning is one of the key capabilities for autonomous humanoid robots. Previous researches have focused on weight balancing, collision detection, and gait generation. Most planners either assume that the environment can be simplified to a 2D workspace or assume that the path is given. In this paper, we propose a motion planning system capable of generating both global and local motions for a humanoid robot in a layered or two and half dimensional environment. The planner can generate a gross motion that moves the humanoid vertically as well as horizontally to avoid obstacles in the environments. The gross motion is further realized by a local planner that determines the most efficient footsteps and locomotion over uneven terrain. If the local planner fails, the failure is feedback to the global planner to consider other alternative paths. The implemented humanoid planning system is an interactive tool that can compute collision-free motions for a humanoid robot in an online manner.

Traditional approaches to the motion-planning problem can be classified into single-query and multiple-query problems with the tradeoffs on run-time computation cost and adaptability to environment changes. In this paper, we propose a novel approach to the problem that can learn incrementally on every planning query and effectively manage the learned roadmap as the process goes on. This planner is based on previous work on probabilistic roadmaps and uses a data structure, called reconfigurable random forest (RRF), which extends the rapidly-exploring random tree structure proposed in the literature. The planner can account for environmental changes while keeping the size of the roadmap small. The planner removes invalid nodes in the roadmap as the obstacle configurations change. It also uses a tree-pruning algorithm to trim RRF into a more concise representation. Our experiments show that the planner is flexible and efficient.

Although virtual humans have been an active research topic for many years, most research has focused on simulating various aspects of a humanoid instead of a crowd of people. We report our progress on generating collision free gross motions for multiple virtual crowds. Each virtual crowd consists of a leader and many followers. The leader is in charge of generating motions for its own group with the motions of other crowds taken into account. The followers use artificial life principles to follow the leader as it moves towards the goal. The high degrees of freedom involved in the crowd simulation system present a great challenge for gross motion planning. We adopt a decoupled path-planning approach, in which the paths being executed by other leaders become the motion constraint of the current leader under consideration. In addition, we take a unified view of search space to extend the planner to consider a whole crowd instead of just its leader. Experimental results show that our planner can efficiently generate satisfactory motions. We believe that such a planning system is a good addition to controlling groups of avatars in a 3D virtual environment.

Research results suggest that one can incorporate motion-planning techniques into the control loop of 3D navigation or tele-operation for more efficient navigation. However, the motion planner with this approach may not scale up well for large workspaces. We propose an approach to overcome this scalability problem. We limit the region of interest for path-finding to a window around the current robot configuration and incrementally update the roadmap in this window as the robot moves. In order to make the roadmap update efficient enough for interactive applications, we adopt a data structure, called rapidly-exploring random tree, to reduce the run-time cost of building the connectivity roadmap. The incremental path planner has been implemented in Java and incorporated into a Java3D-based VRML browser. We compare the performance of this improved planner with the previous one for workspaces of various sizes and analyze the bottlenecks of maintaining such a roadmap. By extending the planning techniques to large work-spaces, we believe that this type of intelligent navigation or tele-operation control will inspire better user-interface design and further researches in planning for large or unbounded worlds.

Due to the rapid evolution of graphics hardware, interactive 3D graphics is becoming popular on desktop personal computers. However, it remains a challenging task for a novice user equipped with a 2D mouse to navigate in an architectural environment efficiently. We think the problem is partly due to the fact that precise navigation control is difficult to achieve with low frame rates. In this paper, we propose a novel approach to improve the effectiveness and efficiency of 3D navigation for architectural walkthrough applications. We adopt a path planner with a probabilistic roadmap to help users avoid unnecessary maneuvers due to collisions with the environment. We modify a Java3D implementation of a VRML browser to incorporate the path planner into the user interface. Experiments show that our implementation of the path planner is very efficient and can be seamlessly incorporated into the navigation control loop. The overall navigation time for traversing a sequence of checkpoints in a maze-like environment can be improved by about a factor of two if the intelligent user interface is used.

After the introduction of VRML, 3D Web browsing has become a popular form of networked virtual reality. However, it is still a great challenge for a novice user equipped with a regular desktop PC to navigate in most virtual worlds of moderate complexity. We think the main problem is due to the fact that a user usually uses a 2D mouse to provide low-level navigation control but the display frame rate is not high enough for this servo loop. We consider an alternative metaphor of allowing a user to specify locations of interests on a 2D-layout map and let the system automatically generate the animation of guided tours in virtual architectural environments. Specifically, we aim to generate animations of customizable tour paths and its associated human/camera motions in an on-line manner according to high-level user inputs. We describe an auto-navigation system, in which several efficient path-planning algorithms adapted from robotics are used. This system has been implemented in Java and adopts common VRML browsers as its 3D interface. We also use the geometric model of our departmental building as an example to demonstrate the efficiency and effectiveness of the system.

We consider the problem of automatically generating viewpoint motions for a virtual camera tracking a moving target. Given the target's trajectory, we plan the motion of a camera to maintain visibility with the target while optimizing certain camera-specific criteria, such as preferred view angle or view distance. Previous work has shown that optimal paths for this problem can be found with a dynamic programming approach. However efficiency for this type of approach, even with a simplified view model, is not good enough for interactive applications. We propose an approach that searches for a collision-free path satisfying visibility constraints in the relative configuration space between the moving target and the tracking camera The search is conducted in a best-first fashion according to time and other optimization criteria. We have implemented this algorithm in a software package that is capable of generating such a tracking motion in fractions of a second. This software module is part of a virtual presence system that generates customizable guided tours to assist users in navigating a virtual scene.

3D collision detection is the most timeconsuming component of many geometric reasoning applications.Any improvements on the efficiency of the collision detection module may have a great impact on the overall performance of these applications.Most efficient collision detection algorithms in the literature use some sort of hierarchical bounding volumes, such as spheres or oriented bounding boxes, to reduce the number of calls to expensive collision checks between polygons.In this paper, we propose an incremental scheme that takes advantage of spatial coherence to improve the performance of this class of algorithms.Experiments have been conducted on a sphere-tree structure for several moving objects.Consistent improvements ranging from 70 to 90 percents were observed.These numbers are actually very close to the theoretical upper bound for such improvements.1.1

Maintainability is an important issue in design where the accessibility of certain parts is determined for routine maintenance. In the past its study has been largely manual and labor intensive. Either by using physical mockup or computer animation with CAD models of a design, the task relies on a human to provide an access path for the part. In this paper, the authors present an automated approach to replace this manual process. By applying results from and developing extensions to research in motion planning and other fields, the authors demonstrate that an automated maintainability study system is feasible. The authors describe general extensions needed to adapt robotic motion planning techniques in maintainability studies. The authors show results from applying such a system to two classes of industrial application problems.

In a dynamic and partially unpredictable environment, robot motion planning must be on-line. The planner receives a continuous flow of information about occurring events and generates new plans, while previously planned motions are being executed. This paper describes an on-line planner for two cooperating arms whose task is to grab parts on a conveyor belt and transfer them to their respective goals, while avoiding collision with obstacles. Parts arrive on the belt in random order, at any time. This scenario is typical of manufacturing cells serving machine-tools, assembling products, or packaging objects. The proposed approach breaks the overall planning problem into subproblems, each involving a low-dimensional configuration or configuration x time space, and orchestrates very fast primitives solving these subproblems. The resulting planner has been implemented and extensively tested both in a simulated environment and with a real dual-arm system. Its competitiveness has been evaluated against an oracle making (almost) the best decision at any one time. The planner compares extremely well.