Calendar - 海角社区

海角社区

Skip to main content

[Defense] Pursuit and Evasion of Drone Swarms and Turrets

Friday, September 30, 2022

11:00 am - 12:30 pm

Daniel Biediger
will defend his proposal
Pursuit and Evasion of Drone Swarms and Turrets


Abstract

The proliferation of inexpensive remotely-controlled drones makes large drone swarms a present reality. Swarms of drones can be useful for overwhelming defensive positions and then loitering in hostile areas near potential adversaries. Even small commercially-available quadcopters can be armed with explosives and can become a significant threat to military forces, as seen in current conflicts. While solutions for countering drones are in development, the problem of countering a swarm of drones falls on existing fixed defensive systems and includes a number of challenges for the defender. The pursuit-evasion problem that arises from the need to engage each of the drones in the swarm in an efficient way and subject to constraints is an example of a Moving Target Traveling Salesman Problem with Time Windows (MTTSPTW). A defending turret must visit each attacking drone in the swarm once, as quickly as possible to counter the threat they pose. This constitutes a Shortest Hamiltonian Path (SHP) through the swarm subject to the time-window constraints of visiting each drone before it can reach the defending turret. Finding this path is made more difficult when the swarm can alter its configuration, changing the relative distances between drones. The rapid pace of engagement, computational complexity, and changing constraints combine to make optimal solvers infeasible. We investigate a mixed strategy of heuristic approaches able to solve the online problem and select a best-effort path through the attacking drones. We determine the optimal theoretical solution for one and two drones and investigate the problem of larger swarms in two and three dimensions. We present an analysis of a critical limitation of existing pan-tilt turrets and the swarm strategy to exploit it. This work presents a comparison of two greedy and two hybrid targeting strategies to counter a swarm, validated by the simulation of random swarms. Using a two-pass approach, with targets grouped by proximity and the path selected by shortest travel distance provided the best results. Once the drones have reached the proximity around a defensive position, they can disable it and loiter in the area. This thesis investigates the bulk movement patterns of the swarm to avoid intra-swarm collisions, while moving the swarm in a hostile area. Using Virtual Reality simulations, it compares the effectiveness of a selection of swarm motion-plans in evading fire from a hostile human user. Parametric knot-like patterns with small random motions provided the best survivability against human adversaries.


Friday, September 30, 2022
11:00AM - 12:30PM CT
F021 (Fleming Building Basement)

Dr. Jaspal Subhlok and Dr. Chang Yun, dissertation advisors

Faculty, students and the general public are invited.