XOTAR’s primary technology focus is the most critical element of enabling general purpose intelligent robots, the robot visual system. XOTAR is researching and developing complex algorithms and software architecture that emulates the known functions, modular visual cortical organization and intermodular communications architecture of the human and higher primate visual system. The Robot Vision system consists of early vision processing for textures, surfaces, and shapes, ultimately indentifying objects and scene organization, monocular and binocular depth perception and active visual processes for selective attention and focus. The company is also developing a very advanced optomechanical vision system with visuomotor and oculomotor apparatus enabling active stereo vision. A reference architecture consisting of algorithms, architectural organization and subsystem communications is designed to be transformed into a custom chipset and embedded systems architecture using an advanced embedded signal processor compilation process, enabling cost effective volume manufacture of affordable, high fidelity robot vision. XOTAR’s strategy is very similar to that taken in the development of perceptual coding architectures that led to video coding for high definition television, image coding for digital photography and audio coding for speech/voice and music processing, all of which led to low cost consumer electronics products enabled with silicon implementations.
Developing a Robot Vision System capable of simulating functions of the human visual system is one of the biggest challenges left unsolved in high technology. Its accomplishment will redefine industrial products across the board, from consumer appliances, vehicles and transportation systems, industrial and agricultural machines, space/defense systems to consumer entertainment products, eventually delivering human inspired robots that provide a wide variety of useful services.
A second primary technology of XOTAR is a prehensile hand – a device capable of grasping and manipulating objects with great dexterity. In primate evolution, the prehensile hand co-evolved with visuomotor control and the adaptive control of eye movement became increasingly more sophisticated as the capabilities of the hand evolved. The Robot Vision system must incorporate the appropriate communication structure and functional partitioning to enable the fine sensing and motor processing technology of the hand to be integrated with the visual system with the appropriate feedback controls supporting the perception-action process; this requires understanding the needs of both subsystems from the onset, especially relative to integrating visual and haptic perception (recognizing objects via touch and integrating that with visual perception of the same object through the cooperating feedback control technology).
Multisensory perception capability will pervade XOTAR’s cognitive systems architecture which will expand to encompass additional sensory modalities in the future, such as vestibular sensing, which augments the vision system for the maintenance of balance, postural stability and spatial orientation along with the spatial mechanisms of the auditory system (hearing ).
Supporting the robot visual system, the prehensile hand and other sensing functions, XOTAR is developing a Perception and Control Law Processor (PCLP) for general robot executive control. The PCLP can be seen as a significant evolution of Artificial Intelligence Inference Engines and Knowledge Representation Systems, which were not designed to integrate perception, decision and control, with the additional requirement that it is based on modal operators (logic) and Bayesian Statistical Inference and utilizes special hardware to support the numeric processing of the soft logic operators. The PCLP serves the combined functions of an integrated central processing complex (massively parallel system), operating system, and language processing engine, providing advanced capabilities for spatiotemporal reasoning, motion and action planning, motor imagery and simulation, and interfaces for lower level visual motor controls. XOTAR’s perception and control law processor architecture is unique in that the core language representation design (intermediate representation) provides a semantic interface between the information perceived by the robot senses and low level natural language semantics. This design will lead to the ability of XOTAR’s robots to communicate in natural language with human collaborators.
XOTAR’s website provides information on our technology, our roadmap, R&D focus, initial markets, management and career opportunities.
