Summaries - Research
Back Defense Against Airborne Swarm Attack
|Division||Graduate School of Engineering & Applied Science|
|Department||Mechanical & Aerospace Engineering|
Dobrokhodov, Vladimir N.
Kaminer, Isaac I.
|Sponsor||Office of Naval Research (Navy)|
We have obtained intelligence on an impending attack by a low flying (to avoid radar detection) swarm of UAVs (lets say 1-10). The intelligence is represented by a probability distribution function (pdf) of an initial time and position of the attack. Regarding the attacking swarm, it is launched in a “fire and forget” manner, i.e. there is no way to trace the origin of the attack. There are three possibilities about their intent:
1) They are piloted by First Person Viewpoint (FPV) and will be manually piloted into a target;
2) They are pre-programmed with a GPS location and will crash into it;
3) They are tracking radar emissions of a NAVY vessel (or any other sensitive installation) and will track them to their origin in a kamikaze attack.
The first two options are the most likely to occur for terrestrial, fixed targets, and would likely involve small numbers of attacking assets. The third would be more likely to focus on naval targets at sea.
The defenders must:
1) Find the swarm. This is exactly the setting of the Herding work we have done before. i.e. this is an optimal search problem in the conditionally deterministic setting. Thus we propose to formulate this problem as a distributed parameter optimal control problem and solve it using SPOC. We will need to address real time issues by optimizing the numerical algorithms currently employed.
2) Neutralize the attackers. Let’s say one of the requirements is that they cannot be shot down. Therefore, we need to investigate using “capture nets” perhaps to scoop them up (may not be an option for smaller size UAVs) or to herd them away from a high impact area to an area where they can be shot down. Here, we need to assume something about the onboard intelligence of the attackers.
A. The attackers are dumb. They simply fly to their destination and do not react to the defenders. Thus here the defenders need to be able to approach and scoop them up. Of course, this requires that defenders have sizable payload capability and have the same numbers as the attackers. Need to be able to do vision based ID with operator OK before finishing the job.
B. The attackers have reactive collision avoidance algorithms, or are being flown by FPV. At first step we learn and identify the type of the collision avoidance algorithm onboard of the attackers. We may be able to use this to our advantage and to try to herd them away from their target to an area where they can be shot down or to have them collide with each other. The approach here again will have to integrate real-time herding trajectory generation, tracking and coordination with vision based estimation
C. The attackers have weapons and try to defend themselves. Thus we are looking at air to air combat. This requires real time distributed trajectory generation that accounts for multiple dynamic constraints of agents. To extend the approach, we can try to evaluate the proximity to optimal solution by resorting to a more formal setting, perhaps as a standard optimal control problem cast in an MPC setting, etc. Lim, Sastry’s ex student, did some work on one on one air combat with a piloted F-18 about 10 yrs. ago.
3) We may also consider low power jamming when we are in the vicinity of the attackers for all types of the attack mission. This could be very effective for FPV, where real-time video and C2 are required for the pilot to navigate. However, jamming will still likely result in a crash, and depending on the location of the attack and the material carried on the enemy aircraft, this may result in unacceptable collateral damage. Thus an airborne capture scheme would be desirable. At sea, jamming might be acceptable but take other forms. Enemy aircraft might not require a C2 link at all, possibly only requiring GPS to loiter in an area until they pick up an RF signal to attack. Jamming in this case might be applied by either jamming GPS while they loiter or spoofing them with an RF signal for them to track so that we can lead them away from our ships
|Publications||Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal|
|Data||Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal|