Organization-based Control of Robot Teams

Type: Project/Thesis

The use of autonomous, heterogeneous robotic teams is one approach to dirty, dangerous and dull missions currently being proposed by DoD. Ideally, we would like to retain human control over these teams to exploit the expertise, critical thinking, and reasoning skills of humans while maintaining the operational efficiency of the autonomous teams. However, there are two major unsolved problems related to this approach: (1) how to control teams of heterogeneous robots in dynamic, uncertain environments, and (2) how to provide this control over a communications channel with a long lag time between commands, execution, and feedback.  In a dynamic, constantly changing environment, robots cannot depend on immediate control from their operators, even if the operator is only a few meters away. In addition, communications may be sporadic.

Currently, most robot teams require multiple operators to carry out a single mission due to the concurrent demands of image management, navigation, systems monitoring and decision making processes. Given this fact, enabling operators to control multiple semi-autonomous robots seems daunting. Thus, we propose to take the focus off the individual robots and center it onto a single entity, the team. While this shift of focus seems minor, it is the key to enabling single operators to effectively control large robotic teams. We plan to investigate how an operator can provide control at the team level instead of individual robot operations.

Thus, the goal of the proposed research is to define a new, high-level control paradigm for cooperative robotic systems based on organizational concepts. We propose to use organizational control techniques to allow single operators to direct large autonomous, heterogeneous robot teams in complex tasks such as team exploration or search and rescue. Our approach will allow operators to modify a team’s organization, thus guiding them in their approach to the solving the problem as opposed to directly intervening in the problem solving process. We hypothesize that controlling the team’s organization will be effective and will help in overcoming problems associated with control in dynamic and uncertain environments, where long command, execution, and feedback lag times may be significant.

To achieve organizational control, we plan to use a four-step process based on our existing team organizational model, which defines and constrains the required elements of a stable, adaptable and versatile team.

1. Define the changes the operator may make to the team’s organizational structure.

2. Define the methods for propagating those changes throughout the team and ensuring the team’s organization remains valid.

3. Evaluate organizational control on simulated teams of robots using the tunnel exploration problem. Simulation allows us to get quick feedback on the team control aspects without having to worry about all the realities of getting it to work on real robots. In addition, simulation will allow us to evaluate the scalability of organizational control using teams of over 100 or more robots.

4. Implemented and evaluate organizational control on a real heterogeneous robot team. While simulation allows us to perform more tests and to stress the limits of organizational control in terms of the number of robots, nothing is more convincing that seeing a new approach work on real, non-trivial problems.

The main result of this research will be a new, high-level control paradigm for cooperative robotic systems based on organizational concepts. Instead of simply supplying high-level goals and relying on team autonomy or providing low-level control, this paradigm will allow the operator to provide high-level goals and then guide the team as whole by influencing its organizational structure in its pursuit of those goals. This research will allow us to discover the benefits of various approaches to organizational control of teams. This research will also help determine when organizational control methods are appropriate.