China
Africa
Explore chapters and articles related to this topic
Dynamic-Size Multiple Population Genetic Algorithms
Published in Ossama Abdelkhalik, Algorithms for Variable-Size Optimization, 2021
This chapter presents a genetic-based intuitive method developed to handle global VSDS optimization problems. Subpopulations, each of them is of fixedsize design spaces, are randomly initialized. Standard genetic operations are carried out for a stage of generations. A new population is then created by reproduction from all members in all subpopulations based on their relative fitnesses. The resulting subpopulations have different sizes from their initial sizes in general. The process repeats, leading to an increase in the size of subpopulations of more fit solutions and a decrease in the size of subpopulations of less fit solutions. This method is called Dynamic-Size Multiple Population Genetic Algorithms (DSMPGA). In space trajectory optimization, for instance, this method has the capability to determine the number of swing-bys, the planets to swing by, launch and arrival dates, and the number of deep space maneuvers as well as their locations, magnitudes, and directions in an optimal sense. This chapter will present the method with several examples on its implementation to the interplanetary trajectory optimization problem. The method can be implemented in other applications as well. The DSMPGA was first published in reference [8].
Flight Planning
Published in Yasmina Bestaoui Sebbane, Multi-UAV Planning and Task Allocation, 2020
Trajectory optimization aims at defining optimal flight procedures that lead to time/energy efficient flights. A decision tree algorithm is used to infer a set of linguistic decision rule from a set of 2D obstacle avoidance trajectories optimized using MILP in [394]. A method to predict a discontinuous function with fuzzy decision tree is proposed and shown to make a good approximation to the optimization behavior with significantly reduced computational expense. Decision trees are shown to generalize to new scenarios of greater complexity than those represented in the training data and to make decisions on a time scale that would enable implementation in a real-time system. The transparency of the rule-based approach is useful in understanding the behavior exhibited by the controller. Therefore, the decision trees are shown to have the potential to be effective online controllers for obstacle avoidance when trained on data generated by a suitable optimization technique such as MILP.
Trajectory adjustment for nonprehensile manipulation using latent space of trained sequence-to-sequence model*
Published in Advanced Robotics, 2019
K. Kutsuzawa, S. Sakaino, T. Tsuji
To avoid the increase in complexity of trajectory optimization, approaches based on machine learning, especially neural networks, have been researched recently. Levine et al. [10] realized various assembly tasks including tight-fitting contact by using neural networks. Yuan et al. [11] realized a planar pushing task with obstacle avoidance by using reinforcement learning. Mordatch et al. [12] realized neural network-based feedback controllers that generate near-optimal walking motions. In addition, neural networks can also be used to reduce computational costs. Although neural networks require high computational costs during training, the trained models are computationally less expensive than most trajectory optimization methods. Zhang et al. [13] showed that neural networks trained by MPC can reduce the computation cost. Similarly, Furuta et al. [14] realized neural networks for dynamic manipulation by copying MPC, and confirmed that neural networks can generate appropriate trajectories faster than the original MPC.
Direct trajectory optimization framework for vertical takeoff and vertical landing reusable rockets: case study of two-stage rockets
Published in Engineering Optimization, 2019
Lin Ma, Kexin Wang, Zhijiang Shao, Zhengyu Song, Lorenz T. Biegler
A direct trajectory optimization framework for the two-phase trajectory optimization problem of VTVL reusable rockets established in Section 2 is presented in this section. The finite-element collocation approach is utilized to discretize the original trajectory optimization problem into an NLP problem solved by IPOPT. A series of complex constraints, such as multi-phase, path and boundary constraints, results in a small feasible region of the discretized trajectory optimization problem. The NLP solver based on Newton’s method often fails to converge; thus, an advanced initialization strategy is proposed to help the solver to overcome the convergence difficulty.
Metaheuristic Algorithms in Smart Farming: An Analytical Survey
Published in IETE Technical Review, 2023
Trajectory optimization is the process of deciding a path that maximize or minimize a criterion of success while meeting a set of given constraints [32]. It is an optimization technique for finding an open-loop solution to an optimal control problem. The application of trajectory-based optimization can be done in computing, where the complete closed-loop solution is not applicable, the taxonomy and time complexity of trajectory-based algorithms is shown in Figure 5 and Table 4 respectively.