A collision-free and connectivity-preserving formation control issue is addressed for a class of nonlinear multi-agent systems (MASs) with external disturbances. The control objective of this paper is divided into two aspects. One is formation control for collision avoidance and connectivity preservation among a leader and communication-connected followers, and the other is obstacle avoidance between followers and potential obstacles including environmental obstacles and communication-connectionless followers. A tangent-type formation error transformation function is introduced for the first topic. With this transformation, an error transformation equation is developed, and the first topic is transformed into an error stabilization issue. Then, an improved continuous differentiable potential function is constructed for the second topic, and the partial derivative of this improved potential function is smooth and continuous as obstacles enter or exit detection ranges. A predictor-based neural network disturbance observer (PNNDO) is developed for generalized disturbances. Prediction errors are introduced to update NNs' weights, and this replacement avoids phenomenon of high-frequency oscillations with overlarge adaptive gains. With a normalized learning method, the number of NNs' learning parameters is also reduced. With the help of the Lyapunov stability, it is proved that the control objective is achieved. The effectiveness of the control strategy is verified by a group of quadrotors.

An optimal tracking control problem is critical for marine surface vessels (MSVs) with full state constraints, input constraints, external disturbances, and model uncertainty terms. For solving an MSV’s position and velocity constraints, a barrier Lyapunov function (BLF) and control barrier function (CBF) are introduced and combined with prescribed performance functions, and then, an adaptive dynamic programming (ADP) optimized tracking control strategy is presented based on the two functions. A dynamic event-triggered mechanism is employed, which reduces the number of triggers by tracking the intermittent transmission of error signals, thus avoiding actuator wear and excessive consumption of communication resources. The experience replay method is brought in the gradient descent method for improvement of the weight learning efficiency. The Lyapunov-based stability analysis shows that the designed tracking control strategy ensures that all signals and weights in the closed-loop system are semi-globally bounded with Zeno-free behavior. Numerical comparison simulation verifies the effectiveness and superiority of the strategy.