China
Africa
Explore chapters and articles related to this topic
Introduction
Published in Peng Hang, Chen Lv, Xinbo Chen, Human-Like Decision Making and Control for Autonomous Driving, 2023
Peng Hang, Chen Lv, Xinbo Chen
The planning algorithms can mainly be divided into five types: (1) graph algorithm, e.g., Voronoi diagram [247] and visibility graph method [277]; (2) heuristic search algorithm, e.g., A star (A∗) [140] and Dijkstra [201]; (3) random search algorithm, e.g., rapidly-exploring random trees (RRT) [36] and probabilistic road map (PRM) [258]; (4) potential field algorithm, e.g., stream function [234], simulated annealing [236], Laplace potential field [193] and boundary value problem (BVP) [45]; (5) spline curve, e.g., Dubins curve [68], Bessel curve [259], B-spline curve [13], sine curve [61], etc. To deal with the optimization issue of motion planning, intelligent optimization algorithms are usually adopted, including Genetic Algorithm (GA) [84], Ant Colony Optimization (ACO) [130], neural network [39], Particle Swarm Optimization (PSO) [150], etc. Though the intelligent optimization algorithms can help plan the optimal path, they rely on infinite iterations to approach a theoretical optimum, namely, the searching process may be too time-consuming due to computational load, which makes them hard to be applied to real-time online planning [129].
Flight Planning
Published in Yasmina Bestaoui Sebbane, Multi-UAV Planning and Task Allocation, 2020
Other approaches can abstract a geometric environment into a topological map based on landmarks. Planning is carried out on this topological map. A planned route has to be converted back to the geometric space for continuous motion control. After obtaining discrete routing information, an admissible and safe path has to be generated for an aircraft to travel from one node to another, compliant with all kinematic and dynamic constraints. Many planning algorithms address obstacle avoidance while planning a path to reach a destination point using A*, D*, Voronoi diagrams, probabilistic roadmap (PRM) or rapidly exploring random tree (RRT) methods. Goerzen in [152] reviewed deterministic motion planning algorithms in the literature from the autonomous aircraft guidance point of view.
Spatial concept-based navigation with human speech instructions via probabilistic inference on Bayesian generative model
Published in Advanced Robotics, 2020
Akira Taniguchi, Yoshinobu Hagiwara, Tadahiro Taniguchi, Tetsunari Inamura
Path planning estimates a path-trajectory from the current position to the goal; however, it is not practical to explicitly obtain the robot with a definitive goal-point each time. Generally, conventional global path planning methods, e.g. the A search algorithm, probabilistic road-map (PRM) [14], rapidly exploring random trees (RRT) [15], as well as extension methods [16], need to specify -coordinates of target-posture or goal-point on the map. There is a limitation in specifying these coordinates in that they depend strongly on the goal-point determination method. Alternatively, recent research on navigational tasks has used vision and language as signals to move toward the target [17–22]. Our method can autonomously determine actions toward the target place from only human speech instructions, using spatial concepts, which are the internal spatial knowledge formed bottom-up.
Towards comfort-optimal trajectory planning and control
Published in Vehicle System Dynamics, 2019
Marlies Mischinger, Martin Rudigier, Peter Wimmer, Andreas Kerschbaumer
A large group of trajectory planners are grid-based algorithms, such as the A* and the D*Lite algorithms. Advanced versions of these algorithms can be used in dynamic environments. However, such grid-based algorithms do not use non-holonomic models for trajectory planning. Another group of trajectory planners are probabilistic algorithms, such as Probabilistic Roadmap and Rapidly Exploring Random Tree (RRT), using a non-holonomic model for the trajectory planning. However, the Probabilistic Roadmap method has a major drawback, since it is based on a graph built in the first planning phase. If there are any changes in the environment, this graph has to be constructed again. This means loss of performance in dynamic environments. In contrast to this, the RRT algorithm does not build a graph. It generates a tree in the whole configuration space beginning from the start point and terminating when the end point is reached. The operations to build the tree are straightforward making this approach very fast.
An approach for solving the three-objective arc welding robot path planning problem
Published in Engineering Optimization, 2023
Xin Zhou, Xuewu Wang, Xingsheng Gu
Welding movement and transition movement make up the robotic movements during the welding process. This article focuses on the transition movement, with the aim of planning as short and collision-free a trajectory as possible from the end of the current weld seam to the start of the next in the workpiece. The planning process can be achieved by a local path planning algorithm instead of by manual operation, which improves the welding efficiency during manufacturing, and reduces the planning time and welding costs. There has been substantial research on local path planning in obstacle avoidance problems, which can be classified into traditional trajectory planning [artificial potential field method (Hou et al. 2019), Dijkstra algorithm (Nowers et al. 2014), A* algorithm (Zuo et al. 2015; Shi et al. 2020)], sampling-based trajectory planning [probabilistic roadmap algorithm (Chen et al. 2021) and rapidly exploring random tree* (RRT*) (LaValle and Kuffner 2000)], heuristic-based trajectory planning [genetic algorithm (Xin et al. 2020), particle swarm optimization (PSO) algorithm (Li and Chou 2018) and ant colony optimization (Akka and Khaber 2018)] and reinforcement learning-based path planning (Gomez Plaza et al. 2009). These algorithms have achieved good results in both academic and industrial areas. However, under the complicated work environment and constraints of the welding robot, the performance of most methods may be limited. To improve the quality of solutions and the search efficiency in complex environments with narrow passages, the potential function-based RRT*-connect algorithm (Wang et al. 2019) was proposed. Informed RRT* (Gammell, Srinivasa, and Barfoot 2014) was proposed to guide the search space for the algorithm after creating a feasible path. Then, an informed RRT* with an improved convergence rate, by adopting the wrapping procedure, was proposed (Kim and Song 2017), resulting in the minimum action of the robot joints. PQ-RRT* (Li et al. 2020) is another improved algorithm, based on the properties of both potential function-based RRT* and quick RRT* (Jeong, Lee, and Kim 2019), where the theoretical proof is given for the completeness, asymptotic optimality and faster convergence, as well as the experimental simulation in a two-dimensional environment. According to the recent research on sample-based path planning algorithms, the RRT* could be improved further through sampling, steering, postprocessing for smoothing, etc.