Towards 3D Path Planning for Free Climbing Robots
摘要
With the increasing capabilities of robotic hardware, the demand for software to fully utilize these becomes apparent. To better utilize the full movement potential, a planning concept for free climbing motions is introduced. For this, classical approaches like rapidly-exploring random trees (RRT) are combined with the versatility of neural networks and reinforcement learning. With this, the traceability of classical algorithms, combined with the versatility of neural networks, can be used to generate complex climbing motion paths.