Dinesh Verma’s research while affiliated with IBM Research - Thomas J. Watson Research Center and other places

Robotic devices have several applications in commercial and defense IoT establishments. However, robots may not always have the capacity to run complex AI based applications, and high speed connectivity to exploit applications in cloud or data centers may not always be present. This situation arises in both defense and commercial contexts, with defense environments lacking sufficient network stability, and commercial environments concerned about data privacy and communications costs. The exploitation of AI capabilities at the edge can enable many use‐cases by bypassing issues with network connectivity. At IBM Research, we have been developing Distributed AI technology and working on several IoT use cases using a robotic dog to enable applications such as visual and thermal inspections to identify anomalous conditions in industrial assets. In this chapter, we will lay out those use‐cases, and discuss key Distributed AI components that enable such IoT applications, and how a robotic environment allows for new capabilities such as re‐positioning a robotic sensor for optimal sensing. We will also discuss relevance of the use‐cases to a military environment.

Ambient sounds provide a wealth of information which can be useful in many IoT solutions in commercial and defense use cases. The application of AI based techniques to traditional acoustic signal processing can provide many interesting use cases which include a diverse set such as detecting faults in industrial manufacturing, identifying possible illnesses in chicken farms, efficient process management of naval facility equipment, and detecting possible intruders at borders. The use of AI provides an augmented capability for a data driven understanding of the environment, but also comes with several challenges. AI models need to operate in environments which may be different from the environment within which they are trained. Effective use of AI models in acoustics requires technologies that can enable these models to retrain themselves, or adapt themselves dynamically within the deployed environment. In this chapter, a system for deploying AI based acoustic models in real‐world environments, and lessons learned from them is described.