AgentScout

AgentScout is an U.S. Army-sponsored research projects developing agent-based coordination and planning techniques for heterogeneous teams in order to comply several kinds of tactical mission types (reconnaissance, convoy protection, surveillance, information collection, logistics, and others).


physics
 

The video begins with a flight over the village. Both the ground and the buildings acts as a obstacles and are used as a collision envelopes used by the physics simulator. The ground assets are simulated by a ray-casted model of a wheeled ground vehicle. In principle, the model can be of any shape and can use any number of wheels. The model is defined using several parameters including wheel friction, suspension stiffness, damping, compression, engine power, break force, and others. Furthermore, the video demonstrates vehicle-to-vehicle collisions and vehicle-to-obstacle collisions.

The fidelity of the physical model helps to anticipate real-world effects for future design of the control, planning, and cooperation mechanisms. E.g. the planning algorithm has to take in account details of the environment as friction, and momentum. The real position of an asset reflects the motion dynamic and introduces the uncertainty of the plan execution which the planner has to be aware of.


drive
 

The scenario shows the maneuver planner and control loops. The UGV is tasked to go through the village from one safe-house to the other. The planner uses maneuvers based on the Reeds-Shepp curves and A* state-space search algorithm. The level of the state space discretization is influenced by a density of the obstacles (the nearer to the obstacle the denser the discretization is). The video shows the density of the state-space by the white dots. The yellow rectangles indicated the area with the denser discretization and the red rectangles shows the no-ride-zones.

Two linear proportional regulators are the main part of the vehicle control loop. One of the regulators controls the steering angle and the other one controls the velocity via engine force and breaking force. The target values of the steering and velocity is obtained from the motion trajectory generated by the planner.


scenario
 

Thisvideo demonstrates a cooperative heterogeneous mission among the UGVs and UAVs. There are two UGVs and three UASs. Initially, the UAVs provides a surveillance of the whole village. The UGVs are tasked to do ground reconnaissance along the path to the safe-house. One of the UGVs is not armed and it is attacked by insurgents. The attacked UGV looses its control and hits a building. The insurgents surround the vehicle and preclude its movement. The UGV calls the UAVs for tighter surveillance of the spot and requests the other armed UGV for help. The other UGV re-plans its route and heads for the position of the first UGV. After the mob is ridden down the UGVs continue to the safe house.


usc-demo
 

In this demonstration scenario, there is a Convoy truck supported by three MDARS UGVs. Additionally, there are two UAVs providing surveillance and tracking of the ground units. The mission is to cross the village from north-west point to the sousth-east point. The village road net is apriori known, but there are also unknown obstacles (red people - insurgents).

Firstly, one of the UAVs is tasked to do a tracking of the convoy to provide close surveillance of the near area, the other UAV is send to the extraction point (end of the mission) to do an area surveillance. The unknown and important crossroads are observed by the UGVs (tasked by the convoy). If a obstacle is found the information is distributed among the ground units and their plans are appropriately altered. Also, the UAVs can observe the ground situation and can find the obstacles.

The UAVs use the effective Zig-zag algorithm performing surveillance over a prespecified rectangular area. The zig-zag pattern is not regular because the algorithm takes into consideration occlusions caused by tall buildings.


usc-tactical-mission
 

The tactical mission begins with a group of ground vehicles (one manned transport truck and four UGVs) supported by UAS platforms. The goal of the mission is to safely go through the village to the objective building on the opposite side, to provide an over-watch during the operation in the building and to ensure safe return of the blue team to the safe-house.

The environment consists of known route graph with buildings (presumably known from satellite reconnaissance) and several a priori unknown entities serving as obstacles for the ground units. There are road blockades (depicted as an icon of a barricade), deteriorated road conditions (depicted as an icon of ground roughness), and rooftop attackers (depicted as red icons of humans). All the obstacles have to be observed both by the UAS and UGVs. The deteriorated road conditions have to be observed by the UGVs explicitly, which means that if a UAS detects such a place it tasks particular UGVs to go to do detailed observation. Additionally, the rooftop attackers can disable a passing UGV. For this scenario, we demonstrate a loss of a vehicle in an area populated by such attackers (the first UGV passing the attackers’ area is disabled and passes all its tasks back to the convoy for future redistribution onto another vehicles). In real case, the rooftop attackers would be detected by the UAS and the UGVs would not be navigated into such areas (in our simulation all other UGVs are informed about the threat and avoid it).

The companion UAS provides two kinds of IRS tasks (a) area surveillance, and (b) target tracking. The surveillance is based on a zig-zag coverage algorithm, which is camera-occlusion aware. The tracking algorithm plans the trajectory of the route in such a way that the airplane circles around the target (since the target’s velocity is expected to be considerably smaller than the velocity of the UAS).

When the convoy arrives to the objective building, the soldiers (depicted as blue icons of humans) are dismounted, and the UGVs take a predefined over-watch positions to cover the neighboring area and the transport moves to the leaving spot. During the operation in the building, several civilians (depicted as a gray icons of humans) and insurgents (depicted as a red icons of humans) leave the building. One of the insurgents is selected as a target for UAS tracking. On the regress route the UGV/UAS surveillance and reconnaissance is reused. As soon as the convoy returns to the safe-house the mission ends.

Top