Table of Contents
Methods #
Graph theory #
- modeling networks of agents, analyze interactions
- sampling-based motion planning with motion primitive automata for multi-modality
The high computational effort, paired with the need for frequent control updates, represents a challenge which I addressed with two approaches. First, I approximated the nonconvex optimization problem using both sequential quadratic programming and a combinatorial reformulation combined with a sampling-based algorithm. Second, to mitigate growing computation times with an increasing number of agents, I designed algorithms acting on a graph representation of the agent network to distribute computations across agents.
Optimization #
- MPC for motion planning and motion tracking
- SCP for fast motion planning
- Graph-based optimization
Learning-based control #
- RL