Active object

Abstract: An interactive table has a display surface on which a physical object is disposed. A camera within the interactive table responds to infrared (IR) light reflected from the physical object enabling a location of the physical object on the display surface to be determined, so that the physical object appear part of a virtual environment displayed thereon. The physical object can be passive or active. An active object performs an active function, e.g., it can be self-propelled to move about on the display surface, or emit light or sound, or vibrate. The active object can be controlled by a user or the processor. The interactive table can project an image through a physical object on the display surface so the image appears part of the object. A virtual entity is preferably displayed at a position (and a size) to avoid visually interference with any physical object on the display surface.

Abstract: A method and system for integrating back-end enterprise applications with Web clients is disclosed. One preferred embodiment comprises a method for invoking an object, comprising the steps of generating a description of the interface of the object; generating metadata representing the interface of the object from the description; storing the metadata; generating a representation of an invocation of the object in a markup language from the metadata; transmitting the representation of the invocation to a client program configured to invoke the object by interpreting the representation; receiving an invocation from the client program; based on the metadata, interpreting the received invocation. In one preferred embodiment, Web clients comprise Web applications generated automatically from CORBA IDL files describing interfaces to objects in the back-end enterprise applications.

Abstract: We study active object tracking, where a tracker takes visual observations (i.e., frame sequences) as input and produces the corresponding camera control signals as output (e.g., move forward, turn left, etc.). Conventional methods tackle tracking and camera control tasks separately, and the resulting system is difficult to tune jointly. These methods also require significant human efforts for image labeling and expensive trial-and-error system tuning in the real world. To address these issues, we propose, in this paper, an end-to-end solution via deep reinforcement learning. A ConvNet-LSTM function approximator is adopted for the direct frame-to-action prediction. We further propose an environment augmentation technique and a customized reward function, which are crucial for successful training. The tracker trained in simulators (ViZDoom and Unreal Engine) demonstrates good generalization behaviors in the case of unseen object moving paths, unseen object appearances, unseen backgrounds, and distracting objects. The system is robust and can restore tracking after occasional lost of the target being tracked. We also find that the tracking ability, obtained solely from simulators, can potentially transfer to real-world scenarios. We demonstrate successful examples of such transfer, via experiments over the VOT dataset and the deployment of a real-world robot using the proposed active tracker trained in simulation.

Abstract: This paper presents a flexible and effective model for object-oriented parallel programming in both local and wide area contexts and its implementation as a Java package. Blending r emoteevaluation and active messages, our model permits programmers to express asynchronous, complex interactions, so overcoming some of the limitations of the models based on message passing and RPC and reducing communication costs.