Noncooperative Herding With Control Barrier Functions: Theory and Experiments

Leader-follower and leaderless bipartite formation-consensus over undirected coopetition networks and under proximity and collision-avoidance constraints

“…In these protocols, all agents reach a global agreement regarding a certain quantity of interest through collaboration [11]. This excludes several scenarios where some agents cooperate, and others compete, as e.g., in the context of herding control [3,7] or in social networks [1]. Networks in which competitive interactions are represented by negative weights on the edges and cooperative ones by positive weights, are called coopetition networks [8].…”

“…In [3], the multi-swarm herding problem is solved under connectivity constraints using a mixed integer quadratically constrained program; in [7], a control strategy is proposed for the non-cooperative herding problem described by first-order dynamics and control barrier functions are used to prevent some agents from escaping from a protected zone. Yet, in [3] and [7], the system is modeled by a network in which all the agents cooperate and the control laws are optimization-based. Moreover, in [7] only a two-agent scenario is considered.…”

“…Yet, in [3] and [7], the system is modeled by a network in which all the agents cooperate and the control laws are optimization-based. Moreover, in [7] only a two-agent scenario is considered. In [6], the bipartite flocking-control problem is studied, and artificial potential functions are used to guarantee collision avoidance and connectivity maintenance.…”

“…This also allows to recast the bipartite formation-consensus problem into the space of the error coordinates. In contrast to [3], [7], and [6], we encode the inter-agent distance constraints using barrier-Lyapunov functions, and our controller is based on the gradient of a barrier-Lyapunov function. With respect to [13], our control law is fully distributed, and we demonstrate our theoretical findings on more realistic scenarios with nonholonomic mobile robots following a leader.…”

Bipartite Formation over Undirected Signed Networks with Collision Avoidance

“…However, there are scenarios in which not all the agents cooperate, but some compete. This is the case, e.g., in Robotics, in the context of herding control [2], [3]. In these coopetition networks [4], competitive interactions are represented by negative weights on the edges and cooperative ones by positive weights.…”