Acceleration-Based Motion Planning for Cooperative Payload Transportation using Multiple Quadrotors

This thesis explores motion planning for transporting a rigid payload using multiple quadrotors connected by cables. While sampling-based planners are attractive for exploring complex environments, they struggle in this setting: effective horizontal motion requires sustained, coordinated drone tilting, yet randomly sampled control inputs frequently lead to instability or violation of attitude constraints. To address this, the thesis first examines two planning approaches that use nonlinear Model Predictive Control (MPC) as a local steering method inside an RRT framework. One samples payload accelerations and tracks them with MPC, while the other samples short position targets. Although both can solve cluttered scenarios, their reliance on repeated nonlinear optimization makes them extremely slow, often requiring hours per environment. Motivated by this limitation, the thesis proposes a new acceleration-based propagation framework that avoids solving an MPC problem at each tree expansion. Instead, payload accelerations are sampled directly and realized through lightweight cable-tension optimization and attitude control. This approach enables fast, dynamically consistent exploration and reduces planning time to under 10 minutes—more than an order of magnitude faster than MPC-based methods. Overall, the work shows that planning in acceleration space provides a scalable and effective alternative for cooperative aerial payload transport, opening the door to faster global exploration and future integration with trajectory optimization.

M.Sc. student under the supervision of Prof. Kiril Solovey.