Adaptive Leader-Follower Formation Control Using $k-\text{WTA}$ and Improved Artificial Potential Fields

Formation control has raised much attention in recent years. In traditional leader-follower model, followers track a fixed leader, which may result in consuming more time and energy when generating formation. It is necessary to explore more flexible control methods. This work studies the adaptive selection of leaders to be followed by followers in leader-follower formation control, in order to achieve the goal of forming a formation as quickly as possible and minimizing the moving distance. A multi-target
$\boldsymbol{k}-\mathbf{winners}-\mathbf{take}-\mathbf{all}$
algorithm (MKWTA) is proposed to select suitable leaders for followers, and an improved artificial potential field (APF) model is proposed to control the movement, collision avoidance and obstacle avoidance of robots. Combining MKWTA and improved APF, it is able for robots to generate formation and move in dynamic environments with obstacles. The effectiveness and feasibility are verified through simulations using differential-driven-wheeled robot.