This work will build two-way information flows between the imaging, mapping, and control systems of a robot, allowing operation in previously inaccessible dynamic and interactive spaces.
Modern machine vision systems generally use one-way information flows to progressively compress visual reality down to the most salient information. In mammalian brains things are much more complex, with substantial information flows from higher-order systems back towards low-level vision.
In this project you will introduce new information loops between low-level sensing, high-level perception, and control. This can include priming neural architectures and driving active sensing systems with high-level context. Fundamental questions arise in representing knowledge, uncertainty, and intentionality inside and outside the robot. The project can also explore the impact of these new information flows on system design, with new possibilities in embodied intelligence and joint design of whole robotic systems. Application areas arise where action is tightly linked to perception including grasping and manipulation, robotic surgery, drone flight in cluttered environments, and autonomous driving in crowded spaces.
Working within the Australian Centre for Field Robotics (ACFR), you will have access to the state-of-the-art robots, facilities, dedicated technical staff, and mentorship available through this world-class research centre. The ACFR undertakes significant field robotics programs in autonomous driving, flight, agriculture, and underwater survey, providing rich opportunities for deployment and validation of novel perception systems.
The opportunity ID for this research opportunity is 2632