 |
 |
 |
 |
| |
|||||
|
|
|
|
|
||||
Securing the Building with Multi-Robot Teams By Marcus daSilva and Andrea Jensenius Imagine you're a squad leader in the middle of a foreign city. You and your squad have just been ordered to clear out the courthouse—a building you think your enemy has abandoned. Rather than go in yourselves and risk the lives of soldiers you command, you each toss a small robot in through a window. Using a handheld computer, you control this mini swarm of semi-autonomous robots, instructing them to explore the building for you. From a safe distance, you and your squad monitor the robots' progress. They communicate with you, build a map as they explore, and share sensor readings like heat, sound, and video. When the wireless communications become unreliable, the robots reorganize to extend their capability. As soon as the robots finish their exploration, you and your squad can safely take control of the building. MITRE is seeking to develop a system of coordinating robots that can determine the best use of both robot and human resources. By making the robots more capable of autonomously acting as a team, the work allows users to be freed from micro-managing the robots so that they can focus their attention on monitoring critical events. Robot Team Historically, many unmanned systems have been controlled through teleoperation: explicit direction given by a remote human operator. Modern systems have introduced some level of autonomy, but the scope is limited and still requires a high degree of human intervention, coordination, and supervision. Some missions, such as reconnaissance and surveillance, can be automated by using multiple unmanned systems working in conjunction to accomplish a task. MITRE's research concentrates on the successful completion of a "building takedown" scenario very much like the one described above. In this scenario, a team of heterogeneous ground robots is charged with searching and mapping a moderate-sized office building and ensuring that it is free of enemy combatants. The robots are deployed with little or no knowledge of the building layout; some spread out, exploring the space and building a map, while others wait behind in anticipation of task assignments. The system is designed to employ each robot for the tasks most suited to its individual capabilities.
Team Strategy Robot team behavior is a result of the interaction that occurs between and among robots as each follows its own policies using local and global sensor information. Each robot possesses individual skills, but the interaction of those skills enables the team to achieve larger and more complex goals than each robot could achieve alone. For example, in a situation such as the building takedown scenario, the explorer robots quickly begin to push the limits of the wireless network and run the risk of losing communications. In the event of such a loss, a second set of robots acts as network relays for the explorers. Each relay positions itself strategically to maximize its ability to connect with an explorer and at least one other node in the network, thus providing a link for the explorer's information to cross larger distances. When one robot is assigned to act as a relay for another, the two become mutually dependent. The relay robot's sole purpose is to maintain connectivity for the explorer robot, and the explorer can travel only as far as the relay permits. Multi-Robot Interface The Multi-Robot Interface (MRI) is a MITRE prototype that allows the sharing of data among multiple robots, as well as with the humans interacting with them. The ultimate goal of the MRI is to provide each robot with an independent interface for data sharing, control, and interaction. The MRI running on each robot in the network is responsible for the regular transmission of sensor data and command messages. In addition, the repository of skills within the MRI is not dependent on a robot's design but rather on its capability. Thus, a skill such as "follow a path" would be the same for a robot with wheels or a robot with tracks. The required capability is only that the robot can move. When a robot joins the network, it shares with the other robots its list of sensors and capabilities, enabling the team to make informed decisions about how to most effectively place resources where they are needed. In this architecture, users are considered to be just another part of the team, and—like every robot on the network—the users have strengths and weaknesses. The human strength in most cases will be the ability to make command decisions on behalf of the team. For instance, if a robot detects something unexpected, perhaps heat, motion, or even voices, it will send a signal to a user asking for further guidance. The user can, in turn, mark the area around the robot as potentially dangerous and request more detailed information. Here, the user is not responsible for tasking robots individually but for sending high-level commands to the team such as "provide video surveillance here." A robot with the right capabilities, in this case a video camera, will then monitor the area. Simulation In addition to providing data sharing capability for humans and robots, the MRI has an integrated simulation component. In a research environment, this provides the ability to seamlessly combine a team of real robots with simulated robots allowing for scalability testing without the prohibitive cost and time commitment of introducing many real robots. In a virtual world mirroring the actual test environment, the simulated robots have the same types of sensors and skill sets as the real robots. To the MRI the real and simulated robots are indistinguishable from one another. In real situations, robots will need to integrate with the other resources around them, whether those are additional robots, human soldiers, or large command systems. Reliability demands that a system for interaction and collaboration be defined in order to ensure mission compliance and to understand how robots will react in complex and hostile environments. The research in this project has given us a glimpse into some of the
strategies that will be essential to effective robot teaming. |
|
||||
| For more information, please contact Marcus daSilva using the employee directory. Page last updated: August 22, 2007 | Top of page |
|||||
Solutions That Make a Difference.® |
|
|
Download this issue [1.1MB]