Multimodal Distributed Perception for Networked Autonomous Systems

The potential of increasingly prevalent networked control systems, including connected automated driving systems (ADS) and networked robots, to enable fast and reconfigurable autonomy underscores the critical need for distributed sensing, perception, and planning architectures that are robust to model uncertainties in time- and safety-critical scenarios. However, perception, scene understanding, reliable state estimation, and safe decision-making for autonomy remain challenging due to computational and processing constraints, environmental uncertainties, and safety concerns. In this talk, distributed state estimation and diagnosis methods based on multimodal remote sensing fused with onboard measurements will be presented. Their performance in terms of computational efficiency and accuracy for reliable robot/vehicle state estimation and safe navigation in highly dynamic scenes and perceptually degraded conditions will also be discussed. Continuous and event-triggered communication between robots and sensor networks, as well as the computational efficiency of the developed multimodal perception system, will be examined for autonomous navigation applications. Finally, recent progress at the Networked Optimization, Diagnosis, and Estimation (NODE) Lab in developing and testing these methodologies for networked robots and intelligent mobility will be presented.