MARS 2020 TEAMS Vignettes
Vignette 1 MissionLab
and CMDLi (Georgia Tech)
In order to support the
fully integrated mission specification capabilities for heterogeneous mobile robots
executed by heterogeneous software platforms, CMDLi (Command Description
Language Interpreter) was integrated into MissionLab. CMDLi parses a text-based
mission specification script as well as an environmental file that contains the
GPS coordinates of known waypoints. A single CMDLi scrip can specify an
integrated mission for multiple heterogeneous robots. During the execution, the
parsed instructions are translated into motor commands in order to carry out
the mission, and each robot notifies its progress to other robots via UDP in
order to support synchronization. Furthermore, CMDLi is also supported by
Player/Stage and ROCI, and thus coordination of the robots executed by
heterogeneous software platforms can be achieved real-time. This demonstration
will cover 1) a brief introduction
of MissionLab, 2) an overview of the CMDLi script, and 3) specification and
execution of a sample CMDLi-based mission in simulation.
Vignette 2
Situational Awareness (UPenn)
Three UGVs (Clodbusters)
and one UAV are tasked to provide situational awareness of a corner of the MOUT
site. The 3 Clodbusters are initially staged at the south corner of the MOUT
site with the situational awareness station. All three robots follow scripted
trajectories visiting a set of prescribed waypoints and collecting panoramic
imagery. These trajectories are determined by a planner that plans trajectories
to waypoints with the stereo obstacle avoider. At the end of this phase the
robots would go to waypoints that are chosen so that they can all talk to the
situational awareness station which are then used to browse the panoramas. The
UAV is tasked to fly over the MOUT site periodically taking a sequence of
pictures at prescribed waypoints and transmitting these to the base station.
The user accesses these pictures to acquire a database of images. This allows
us to show a sequence of images taken from approximately the same location for
facilitating change detection.
Vignette 3 3D Mapping
and Autonomous Navigation (USC)
Using the Segway RMP carrying
a laser rangefinder, we will demonstrate autonomous navigation in the MOUT site
with minimal direction from the user (in the form of a goal point). The robot
will plan a path to the goal, and execute the path. The data from the laser
scanner will be used to build (offline) a 3D map of some of the buildings in
the MOUT site. We will demonstrate simultaneous localization and mapping
(offline) algorithms based on laser rangefinder scans from Segway RMP,
Autonomous navigation/obstacle avoidance through urban environments based on
solely on goal point specification, and dynamic path planning in real time.
Vignette 4 Radio
Mapping (UPenn)
3 UGVs plan a set of
trajectories (waypoints) to map the radio signal strength between pairs of
points in the MOUT site. As the robots traverse the site a radio map is built
in real time. A map showing radio signal strength between robot locations is
shown on a desk top. The utility of this is demonstrated by sending an image
across a string of robots to a base station that a human operator can view.
Vignette 5
Cooperative target search, identification and localization (UPenn)
The UAV is tasked to fly
a search pattern overhead using the cameras to spot for an appropriately
colored target. When that target is spotted the estimates for the target
coordinates are relayed to a UGV which traverses the MOUT site searching the
approximate area indicated by the UAV. Overhead cameras provide information to
the UGV to improve its position estimate. A desktop display will show the error
ellipsoid associated with different estimates. Meanwhile, the UGV uses its
cameras to confirm the existence and positive identification of the
target.
Vignette 6 Mission
execution while maintaining communication constraints (Georgia Tech)
(a) Communication
Recovery for Multi-Robot Teams
This demonstration will
show an example of communications-sensitive robot behavior in the context of
communications recovery. A team of
robots originating at a base station in a hummer will perform a reconnaissance mission in which communications
constraints between the team members cannot be met. Once the communication constraints have been broken, the
robots will recognize the failure and initiate communications recovery
behaviors to re-establish communication with their other team members. Once
communication recovery has been successful and the team is once again in radio
contact with each other, the team will initiate a contingency mission to gather
its members at an alternate waypoint and then return to the base station.
(b) Value-Based Communication Preservation
Demonstration
Information is desired
about the buildings in the southeast corner of the village. The robot operator deploys two robots
from the Hummer. Robot #1 is
tasked with exploring the area in question by running through a set of GPS
waypoints. To maintain data flow
from robot #1, robot #2 is tasked with preserving communication between robot
#1 and the Hummer using Value-Based Communication Preservation (VBCP). VBCP uses a rough connectivity model
and the locations of the cooperating agents to guide the robot to areas of
better connectivity. As robot #1 moves along the pink building and up the
alley, robot #2 responds to robot #1's actions by moving to positions that
maintain communication between robot #1 and the Hummer. Just as robot #2 reaches the end of the
alley and seems about to disappear around the corner of the yellow building,
robot #1 moves to the other side of the alley to prevent this occlusion. When robot #1 is finished with its
exploration, it makes its way back toward the starting point. Robot #2
continues to maintain communication, moving with robot #1 back toward the
Hummer.
Vignette 7 3D Air
Ground Coordination: Cooperative Localization and Tracking (USC)
Using the AVATAR
helicopter robot, equipped with a full inertial package, and a downward
pointing camera, we will demonstrate autonomous aerial navigation in the urban
canyon environment. The camera imagery will detect and localize ground robots.
We will demonstrate autonomous aerial navigation in urban canyon environments,
detecting and tracking ground vehicles from the air using a camera, thus
ocalizing ground robots.
Questions
For directions,
logistics of ingress and egress, etc. contact
(a) Irv Rodriguez at rodriguezi@benning.army.mil, phone: 706.545.5109, mobile: 706.575.8421
(b) Mike Kennedy at kennedym@benning.army.mil,
phone: 706.545.9215, mobile: 706.575.8181.
For questions about the
MARS 2020 programs or demos, contact Vijay Kumar at kumar@cis.upenn.edu,or
call 215.898.3630 or 215.898.0374.