Motion Planning
This chapter presents a kinodynamic motion planner for computing agile motions of quad-rotor-like aerial robots in constrained environments. Based on a simple dynamic model of the UAV, a computationally-efficient local planner is proposed to generate flya
- PDF / 720,502 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 21 Downloads / 206 Views
Abstract This chapter presents a kinodynamic motion planner for computing agile motions of quad-rotor-like aerial robots in constrained environments. Based on a simple dynamic model of the UAV, a computationally-efficient local planner is proposed to generate flyable trajectories of minimal time. This local planner is then integrated as a key component for global motion planning using different approaches. The good performance of the proposed methods is illustrated with results in simulation, as well as a preliminary experimentation with a real quad-rotor.
1 Introduction When planning motions for Unmanned Aerial Vehicles (UAVs) such as quadrotors, it is important to consider the dynamic model of the system since not every geometrically valid path corresponds to a feasible motion. For example, because of its dynamic behavior, flying upside down, even for a relatively short period of time, is hardly manageable for a fixed-pitch quadrotor. Aiming to avoid the difficulties and the high computational cost involving kinodynamic motion planning [1], the problem is usually treated in two stages. The first stage applies a basic path planning approach, disregarding dynamic considerations. For this, sampling-based path planning algorithms [2], such as the Rapidly-exploring Random Tree (RRT) or the Probabilistic Roadmap (PRM), can be used to produce a collision-free path for the center of mass of the robot. Indeed, since the robot orientation depends on dynamic aspects, collisions cannot be tested for the robot itself but for its smallest bounding-sphere. In a second stage, this path, usually consisting of a sequence of straight-line segments in A. Boeuf (B) · J. Cortés · T. Siméon LAAS-CNRS, Université de Toulouse, CNRS, Toulouse, France e-mail: [email protected] J. Cortés e-mail: [email protected] T. Siméon e-mail: [email protected] © Springer Nature Switzerland AG 2019 A. Ollero and B. Siciliano (eds.), Aerial Robotic Manipulation, Springer Tracts in Advanced Robotics 129, https://doi.org/10.1007/978-3-030-12945-3_23
317
318
A. Boeuf et al.
Fig. 1 Solution to the slot problem obtained with a basic RRT algorithm using the proposed local planner as a steering method
R 3 , has to be transformed into a dynamic trajectory. Trajectory generation methods, such as [3–5], can be applied in this stage to each portion of the path. The overall trajectory then consists of a sequence of movements from one hovering position to another, which leads to severe sub-optimality in terms of execution time. However, such an unsuitable trajectory can be subsequently optimized using several types of algorithms. The aforementioned decoupled approach is computationally efficient, and can be successfully applied to solve many motion planning problems for UAVs (see for example [6–8]). However, several classes of problems cannot be treated using this approach, because the robot orientation cannot be properly considered at the geometric stage. One of such problems, which we refer to as the slot problem, is illustrated in Fig. 1. In this problem, the
Data Loading...
