SDF-Based Reinforcement Learning for Adaptive Path Planning and Formation Control of Multiagent Systems

Formation control based on path planning is an important and critical research topic in robotics, which focuses on generating collision-free paths for multiagent systems (MASs) from an initial position to a target position while maintaining the desired formation. This article realizes adaptive path planning and formation control for MASs with the presence of lumped uncertainties and saturation input. To achieve this goal, a hierarchical adaptive formation planning and control (HAFPC) framework, including a formation path planning layer and an adaptive formation control layer, is constructed. In the formation path planning layer, the signed-distance-field-based formation path planning (SDF-FPP) algorithm is proposed to find a collision-free continuous trajectory in an unknown environment from the initial position to the target position. Based on this collision-free trajectory, a nonanalytic function that evaluates the shortest distance between this collision-free trajectory and obstacles is computed via the signed distance field (SDF) method. Then, this nonanalytic function will be further processed in the next layer for obstacle avoidance of all agents. In the adaptive formation control layer, the proposed adaptive-offset formation control (AOFC) algorithm converts the nonanalytic function into the adaptive offset functions for all agents and manipulates MASs to achieve adaptive formation control for obstacle avoidance with the presence of lumped uncertainties as well as saturation input. Simulations are presented to validate the proposed architecture.