Coordinated Autonomy through Dec-MCTS with Communicated Behavioural Intent

Project Member(s): Best, G., Fitch, R.

Funding or Partner Organisation: University of Adelaide (University of Adelaide funding source)
University of Adelaide (University of Adelaide funding source)
Defence Science and Technology Group of the Department of Defence (Defence Science and Technology Organisation of the Department of Defence)

Start year: 2022

Summary: This project will develop new representations of behavioural intent for low-latency coordinated behaviour in a team of robots performing Dec-MCTS. These new representations will achieve four objectives: (1) accurately portray the behavioural intent of a robot, as modelled by the current Dec-MCTS plan, (2) be in a concise form suitable for systems with limited communication bandwidth, (3) are interpretable by receiving robots while updating their own actions to satisfy the team’s goals, and (4) come with certificates verifying robust performance and safety. The expected outcome is the capability to deploy Dec-MCTS in scenarios where the robots are strongly required to adapt their actions based on changes in the environment and the intent of other robots in the team, and high assurance is required. These intent representations can also be generalised to model the behaviours of adversarial and neutral agents in the environment. Another benefit of these new representations will be improved introspection of the coordination behaviour by human-on-the-loop supervisors due to the human readability of such policy representations. These capabilities and benefits will be demonstrated through improved coordination in defence-relevant scenarios, such as scouting and reconnaissance missions, and will complement the CIs’ other ongoing defence-funded research projects.

FOR Codes: Control engineering, mechatronics and robotics, Emerging defence technologies