SRI's coverage planning algorithms are encapsulated in a spatial reasoning system called Spare. This system was developed during the SRI Centibots Project, under the DARPA Software for Distributed Robotics (SDR) program (Contract: NBCHC020073). Spare was successfully demonstrated as part of the final Centibots system, in which 100 robots were deployed to explore and secure large unknown environments [Ortiz et al. 05].
The Spare system combines a map of the environment with a mission specification to determine an allocation of spatial goals and particular behaviors to each robot. Each mission is complex and exhibits a set of constraints. For example, in the perimeter search mission, robots must exhaustively cover the area of interest while taking into account the limited range of their FIDO sensor. Simultaneously, they must maintain a radio contact (possibly through relays) to the OCU. Certain areas may require more attention (and thus more time) than others, which should also be taken into account during planning. For instance, building perimeters and vehicles, as well as operator-designated zones of interest, will likely require finer search granularity than large open areas.
Spare optimizes the placement and behavior of the robots in order to maximize the overall multi-criterion utility, taking into account factors such as distances between robots, energy usage, line-of-sight maintenance, and zones of special interest. Because Spare can optimize across multiples constraints, the commander is able to adjust the relative importance of each constraint when specifying the mission.
Once the commander has defined the mission, solution generation proceeds in two phases. First, Spare extracts an abstract discrete representation of the map. Second, it uses optimization techniques to generate the best allocation of robots with respect to the abstract map and the mission constraints. The allocation is then handed off to the robots for execution.
The Idaho National Laboratory (INL) has developed an Intelligence Kernel architecture that has been ported to a wide variety of mobile robots [Pacis, et al. 04]. The Intelligence Kernel includes capabilities for path planning and obstacle avoidance as well as a graphical user interface that can be used to interact with multiple robots.
One application of this architecture has been in mine detection, and this system has been selected by the US Army for the Autonomous Robotic Countermine System (ARCS) project. In one set of trials, a UGV and a UAV using the Intelligence Kernel were able to cooperate in a mine detection task, with the UAV identifying potential threats and the UGV moving in to examine those threats. This system was able to locate the position of mines and display those positions on an aerial map of the terrain (Figure 13).
During the Fireseeker Project we plan to integrate the Wayfarer urban navigation system with the INL Intelligence Kernel and graphical user interface. This will allow multiple Fireseeker robots to display the position of suspected IEDs on a single Operator Control Unit (OCU) interface display, superimposed onto aerial or satellite imagery. The terrain images may be provided from a pre-existing database or generated in real-time from UAVs.
Figure 13: Aerial map of mine locations generated by UAV/UGV team
Integration of Wayfarer urban navigation capabilities with the INL Intelligence Kernel will also allow greater interoperation with other robots using this architecture. These include systems developed by the Army Night Vision Laboratory (NVL), the Navy Space and Naval Warfare Systems Command (SPAWAR), NASA’s Johnson Space Center, and Carnegie Mellon University (CMU). This architecture fully supports the Joint Architecture for Unmanned Systems (JAUS) and will allow Fireseeker to interoperate with JAUS-compatible robots and OCUs.
Share with your friends: |