Cooperative Control of Multi-Quadrotors for Transporting Cable-Suspended Payloads: Obstacle-Aware Planning and Event-Based Nonlinear Model Predictive Control

TL;DR

This work tackles the problem of coordinating $N\ge 3$ quadrotors to transport a cable-suspended payload in cluttered, dynamic environments. It fuses global A* path planning with local nonlinear model predictive control (NMPC) and an event-triggered update mechanism that relies on a perception stack combining multi-camera SLAM/VIO with event-camera data, augmented by a Graph Neural Network (GNN) for processing asynchronous events. Key contributions include (i) an event-triggered NMPC with obstacle-aware planning, (ii) a perception and SLAM framework that fuses multi-camera VIO and event data for robust localization and mapping, (iii) dynamic waypoint replanning via A* as obstacles change, and (iv) extensive simulations showing improved energy efficiency, computation, and responsiveness. The results demonstrate safe, efficient, and scalable cooperative payload transport, with practical impact for construction, delivery, and inspection missions where dynamic obstacles and limited onboard resources are critical considerations, underpinned by models with $m_i$, $R(\Theta_i)$, $T_i$, and 6-DoF payload dynamics represented within the NMPC framework.

Abstract

This paper introduces a novel methodology for the cooperative control of multiple quadrotors transporting cablesuspended payloads, emphasizing obstacle-aware planning and event-based Nonlinear Model Predictive Control (NMPC). Our approach integrates trajectory planning with real-time control through a combination of the A* algorithm for global path planning and NMPC for local control, enhancing trajectory adaptability and obstacle avoidance. We propose an advanced event-triggered control system that updates based on events identified through dynamically generated environmental maps. These maps are constructed using a dual-camera setup, which includes multi-camera systems for static obstacle detection and event cameras for high-resolution, low-latency detection of dynamic obstacles. This design is crucial for addressing fast-moving and transient obstacles that conventional cameras may overlook, particularly in environments with rapid motion and variable lighting conditions. When new obstacles are detected, the A* algorithm recalculates waypoints based on the updated map, ensuring safe and efficient navigation. This real-time obstacle detection and map updating integration allows the system to adaptively respond to environmental changes, markedly improving safety and navigation efficiency. The system employs SLAM and object detection techniques utilizing data from multi-cameras, event cameras, and IMUs for accurate localization and comprehensive environmental mapping. The NMPC framework adeptly manages the complex dynamics of multiple quadrotors and suspended payloads, incorporating safety constraints to maintain dynamic feasibility and stability. Extensive simulations validate the proposed approach, demonstrating significant enhancements in energy efficiency, computational resource management, and responsiveness.

Figures (3)

• Figure 1: The control block diagram of the proposed approach
• Figure 2: Cable-suspended load transportation by MAVs
• Figure 3: Event-triggered nonlinear model predictive control with proposed obstacle-aware planning for position trajectory of quadrotors in 3D in the presence of static and dynamics obstacles