Modification of Kohonen Rule for Vehicle Path Planing by Behavioral Cloning
The problem of path generation for the autonomous vehicle in environments with infinite number obstacles is considered. Generally, the problem is known in the literature as the path planning. This chapter treated that problem using the algorithm, named MKBC, which is based on the behavioral cloning...
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Capítulo de libro |
| Lenguaje: | inglés |
| Publicado: |
IntechOpen
2008
|
| Materias: | |
| Acceso en línea: | https://www.intechopen.com/chapters/5363 |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | The problem of path generation for the autonomous vehicle in environments with infinite number obstacles is considered. Generally, the problem is known in the literature as the path planning. This chapter treated that problem using the algorithm, named MKBC, which is based on the behavioral cloning and Kohonen rule. In the behavioral cloning, the system learns from control traces of a human operator. Kohonen rule connected with the weighting coefficients, while the MKBC algorithm does not use the weighting values as values from the previous time, but permanentlly uses the training values as weighting values. That is something which enables an intelligent system to learn from the examples (operator's demonstrations) to control a vehicle in the process of the obstacles avoiding, like the human operator does. Like that, the very important MKBC characteristic is the symplicity. The MKBC simplicity is something which is so obviously, specialy according to the RBF neural network and the machine learnig algorithm which is used the previously. Following the MKBC given context the problem narrow passage avoiding and the goal position reaching fundamentally is observed. Namely, defining if ? then rule, according to the named cases is treated as destroying of the consistency of the methodology. In that sense, using MKBC neural network the solution was found. A the end, the autonomous vehicle mathematical model which is given by nonlinear equations describing a 12 state dynamical system is used and in that case the MKBC algorithm is applied successfully. Eventually, as it has been illustrated the previously, the advantage of the entire methodology lies in the fact that a complete path of the vehicle can be defined off-line, without using sophisticated symbolical models of obstacles. These are facts that MKBC algorithm and the given methodology substantially differ from the others. In the next phase it is expected to confirm results in on ? line simulation process. Key words: vehicle path planning, behavioral cloning, cloning success, obstacle avoiding, machine learning, Kohonen rule, neural network, Shark dynamical model. |
|---|---|
| ISBN: | 978-953-7619-01-5 |
| DOI: | 10.5772/6009 |