Nonlinear Distributed Model Predictive Flocking with Obstacle Avoidance

In this paper, we present a framework for nonlinear distributed model predictive flocking with obstacle avoidance, the pursuit of group objectives, and input constraints. While most existing predictive flocking frameworks are only applicable to agents with double-integrator dynamics, we propose a general framework for nonlinear agents that furthermore allows for the independent tuning of cohesive and repulsive inter-agent forces. To reduce the computational complexity, the resulting nonlinear program is solved as a sequential quadratic program with a limited number of iterations. The performance of the proposed algorithms is demonstrated in simulation and compared to a non-predictive flocking algorithm.