Distributed Trajectory Planning for Multiple Autonomous Aerial Vehicles

In my Bachelor’s thesis, I developed centralized and distributed trajectory planners for multiple autonomous aerial vehicles and compared their performance.

With an increasing number of vehicles featuring some degree of autonomy, possible interactions between agents must be taken into account to guarantee safe operation within a confined space. Distributed planning is one way to tackle this: instead of a single central planner computing trajectories for everyone, each agent solves part of the problem itself.

I first developed a centralized planning algorithm that generates safe trajectories for multiple CRS EmboRockETH agents. I then cast this planner into distributed form and, based on the distributed optimization problem, implemented decentralized planners using ALADIN and round-robin scheduling.

The performance of all trajectory planners (centralized, distributed and the more naïve decentralized variants) was evaluated and compared using various problem setups.