Defense Science & Technology Laboratory
About the Project
This project will demonstrate a fault tolerant autonomous robotic system that is able to continually detect, in real-time, changes in the air environment, and construct a hazard map of potential threats.
At the same time the system will have an autonomous control and navigation system implemented using a state of the art approach based on self-repairing neural networks. These networks go way beyond traditional paradigms by including astrocyte cells which have recently been modelled to capture the repair capability exhibited in the human.
This will afford a resilience to various types of potential failure within the controller; resilience of the controller is initially proposed as way of demonstrating capability. Human operators will be able to provide real-time feedback to the performance of the system to allow for machine learning to improve the overall performance of detection and navigation through a reinforcement based approach to learning.
This proposal will deliver a proof of principle demonstrator of a fault-tolerant autonomous robotic system, capable of mapping hazards chemical environments and identifying key hazards of interest.
Human users will be able to have real-time feedback of the hazard map and items identified, as well as perform limited control of the unit to compliment the units autonomy. The work is based on previous research, by the investigators, on self-repairing neural networks, real-time anomaly detection and robotic deployment.
Projected key outcomes
- Delivered a proof of principle bio-inspired algorithm that demonstrated brain inspired self repair
- Delivered a proof of principle demonstrator of a fault-tolerant autonomous robotic system, capable of mapping hazards chemical environments and identifying key hazards of interest