Real-time (RT) sensing with enhanced video capabilities is a powerful tool capturing a significant variety of data, thus providing an excellent substrate to build up an accurate context abstraction. Machine capability of creating a good context abstraction will boost several applications. 3D video allows the recognition of objects shapes, human gestures and facial expressions, improving human-to-machine interaction. Surveillance systems, empowered by automatic recognition of gestures, will suddenly detect a threat. Facial expressions may reveal the status of elderly or disabled people. Remote driving and decision making process of robots for scientific or rescue expeditions will improve. VISION aims to developing an innovative infrastructure providing RT sensing services, with particular emphasis on 3D video, with mobile and context-aware operation. Hence, it will address the numerous challenges raised by the limitations of current technology for wireless sensor networks. VISION will exploit the 60 GHz UWB radios enabling broadband transmissions, miniaturized devices and reduced interference. Due to the inherent high resource consumption of broadband wireless necessary for RT 3D video, VISION sensor network will be optimized at all layers. A comprehensive channel model, based on propagation measurements, will be derived and used to design the UWB signal waveform for reliable broadband transmission. Novel techniques to manage the huge number of nodes required by ubiquitous sensing, and innovative tools to support the development process of intelligent services will be designed. Full cross-layer adaptability to external conditions will assure the system be able to manage the available resources to provide the best achievable quality of service guaranteeing graceful degradation for video, audio and sensing applications. Innovative sensor nodes based on the above concepts will be designed and prototyped for the VISION demonstration.