Agile micro aerial vehicles will have a massive societal impact over the next decades, creating novel opportunities for large-scale precision agriculture, fast delivery of medical supplies, and disaster response, and providing new perspectives on environmental monitoring and artificial pollination. This future requires the design of robust and lightweight perception algorithms, which interpret sensor data into a coherent world representation, enabling on-board situational awareness and decision-making.
In this talk, I present my work on robust and lightweight robot perception, including the design of algorithms for fast visual-inertial navigation and the development of the first certifiably correct approach for localization and mapping. I also discuss the challenges connected to scaling down perception to nano and pico aerial vehicles, where sensing and computation are subject to strict payload and power constraints. I argue that enabling autonomy on miniaturized platforms requires a paradigm shift in perception, sensing, and communication, and discuss how we can draw inspiration from nature in designing the next generation of flying robots.