As robots progress out of the laboratory and into civilization, it becomes important to ensure that they perform in ways consistent with our ethical expectations. This applies to robotic domains ranging from the battlefield to the household. In this talk, specific robot architectural design recommendations are presented for (1) post facto suppression of unethical behavior through the use of an ethical governor, (2) the use of a behavioral design methodology that incorporates ethical constraints from the onset, (3) the use of affective functions that serve as an adaptive component in the event of unethical action, and (4) a mechanism in support of identifying and advising operators regarding their ultimate responsibility for the deployment of such systems.

Without loss of generality, specific examples are drawn from the potential use of lethal force by autonomous robots developed for the military. As weaponized robotic systems are being introduced into the battlefield at an ever increasing pace, the consequences of this technological progress need to be examined carefully and presently. In this talk, I outline the philosophical basis, motivation, theory, and design recommendations for the implementation of an ethical control and reasoning system potentially suitable for constraining lethal actions in an autonomous robotic system so that they fall within the ethical bounds prescribed by the Laws of War and Rules of Engagement, which serve as internationally agreed upon ethical norms. Results obtained to date are surveyed. The implications and the generalizability of this research for other robotic domains is also considered.

21st century has brought an abundance of cheap sensors, from surveillance video and infra red cameras to radar on the chip, to small accelerometers and microphones. New sensor concepts including differential time and differential doppler are becoming feasible. Equally important is the progress in computational technology which often provides cheap and ample computational and memory capabilities even in embedded and autonomous systems. Object tracking can now enhance existing and open a plethora of new applications.
Object Tracking in clutter applications now include traffic monitoring and control, smart cars, asset tracking, crowd monitor and control, robotics, emitter geolocation, speaker tracking, wireless sensor tracking, UAV (Unmanned Arial Vehicle) based video and emitter tracking, etc.
Object tracking based on Bayes formula enables easy to understand, easy to implement, easy to maintain solutions. A suite of algorithms exist, which enable performance/computational requirements trade-offs. Measurements from multiple sensors may be fused for enhanced capabilities.

As Digital Video Surveillance Systems become common place, customers are installing 100's and 1000’s of cameras for enhancing security and improving operations in their facilities. The challenge of monitoring the cameras using human vigilance is known to be ineffective and cost prohibitive. Customers are increasing using smart surveillance systems or video analytics technology to automate the monitoring of cameras. Smart surveillance systems use computer vision and pattern recognition technology to automatically alert operators to suspicious behaviors and to allow rapid search of video events. This talk presents the enabling technologies for smart surveillance, object detection, tracking, object classification and video indexing. These technologies are presented in the context of an end to end system architecture which goes from video processing to alerting and searching interfaces. The talk will present applications of smart surveillance in both city surveillance and retail environments. The talk will draw upon real world deployments of IBM's Smart Surveillance Solution.

Multifinger manipulation is considered to be the cornerstone that triggered human evolutionary development towards intelligent behaviour, once bipedalism made it possible by freeing hands for manipulating objects. Indeed, most of our physical interaction with the world is mediated by the use of our hands. Cognitive robotics approaches to robotic manipulation are however rare. An intelligent service robot will be severely impaired if it is not able to reliably perform simple manipulation tasks in everyday environments such as opening a door to enter another room, or pulling a drawer open to take something out. Today's robots exhibiting such abilities do it in an ad-hoc fashion with very precise models of the environment. This is in evident contrast with the versatility of the manual skills readily apparent in primates. Cognitive Neuroscience supports that one of the key mechanisms for manual actions relies on the multisensory interplay between vision, proprioception and touch. The motor system in the cerebral cortex is not only involved in motor functions, but also plays a role in cognitive functions such as sensorimotor transformations, action understanding, or decision making about action execution. It acquires knowledge about the external world through sensorimotor experience and this knowledge is not static: motor action development and learning occurs through plastic reorganization of movement representations, in a way that reflects the kinematics of acquired skilled movement, encoding the memory for motor experience. In this talk I present a multidisciplinary perspective based on sensorimotor integration for multifinger robot manipulation in which well-established robotics techniques and hardware go hand in hand with new systems design and engineering principles based on findings in neuroscience and psychophysics.