Robotic Survey Group Proposal

One critical need for autonomous mobile robots is localization. Self contained approaches, such as the use of encoder data, are error prone, and can quickly lead to significant errors in position and orientation estimation. As a result landmark-based localization has be used to allow mobile robots to establish their position in well known environments. While these environmental cues can be used as location markers, there is no certainty that they will either exist, or be well distributed for the localization task. Recent research into autonomous mapping of unknown or ill-defined domains has focused on the establishment of fiducial marks, which are then used by the mobile robots to determine position. Such marks are placed into the environment, and their locations are well established and disseminated to the mobile robots. Thus the robots can determine their positions and orientations relative to the fiducial marks, and their absolute postion in the domain.

In well defined, static domains this task is straight-forward. Off-line calculations can be done to determine both the optimum number and the optimum placement of such localization markers to cover a known area. If these markers are passive, as many as needed can be placed and registered to cover an area. However, in the event of a clean-up after a catastrophic, hazardous release, such passive, pre-registered marks can fail. There may be insufficient light to make them visible, the geometry may have changed to obscure them, or they may be covered with, or damaged by, the very material that needs to be cleaned up. In all of these cases, human operators must enter the area, exposing themselves to the hazards, to either install new location markers, or clean up the spill. Also, as the cleanup evolves, different spatial areas will be identified, accessed, cleaned, and then the work area will move on, requiring the ongoing exposure of humans to maintain the localization markers.

One alternative is to augment the cleanup robots with the survey equipment needed to perform self-localization. This approach has benefits in that a more homogeneous RIM team can be deployed into an area, and all robots would be interchangeable. However, the cost of high resolution survey sensors, and survey specific inter-robot distance sensors would significantly impact the 'deployment cost' of the cleanup robots. In addition, the cleanup task itself is a complex task that may stress the computational resources of the robots. The survey task is also computationally complex, and time consuming. With separate survey robots, the expansion of the surveyed area can proceed while the cleanup of already surveyed areas is being completed. By providing robotic 'specialists', the cleanup robots can be designed to focus the resources they have available (power, weight, computational capacity) onto the cleanup tasks, while the survey robots can focus resources on providing a stable localization infra-structure to any robotic team that requires high quality position and orientation information in partially engineered environments.

As an alternative to human exposure, a semi-autonomous robotic survey group (RSG) could be deployed.The survey robots will be based on the Pioneer AT all-terrain platform, augmented with additional sensors and survey locator beacons. These robots have shown themselves to be robust, and capable of traversing rugged terrain, and can provide a stable platform for the survey tasks.

The RSG of semi-autonomous mobile robots would enter a contaminated area, survey it and provide active, relocatable localization beacons for the robotic cleanup team. The survey group would be capable of building a map of the area, and after analysing the geography of the current area of interest, would deploy itself to provide near optimum localization signals to the other robots. As the work area evolved, the survey team would redeploy to provide continuing coverage of the area.

We propose to develop a survey group that consists of a small number (4-6) of all-terrain semi-autonomous robots, with high resolution onboard localization equipment. In addition, each would carry a laser-based beacon that would emit a continuous data stream of angular displacement, and global location. Thus any worker robot entering the work area could recieve a signal indicating its own position relative to the beacon, and the absolute postion of the beacon. Upon receipt of two such signals, the absolute position and orientation of the worker robot can be determined, and with additional signals precise error analysis can be done.

If the spill has resulted in new obstructions in the environment, the RSG can produce accurate occupancy grids as these changes are discovered, preventing problems with the worker robots. In addition, if deployed into an unknown environment, the RSG can produce maps which can detail both obstacles and accessibility grids, allowing the effective routing of other, limited terrain robots.