Safe collision avoidance for self-driving vehicle chassis within electric mobile assembly
View/Open
safe-collision-avoidance-for-self-driving-vehicle-chassis-within-electric-mobile-assembly-final.pdf (3,474Mb) (Restricted access)
Document typeMaster thesis
Date2019-09-20
Rights accessRestricted access - author's decision
Abstract
Electric mobility is a challenging research field nowadays, mainly as a result of high production costs. A need for flexibility is clear, being a self-driving vehicle chassis an encouraging solution. State of the art navigation techniques, like the ones employed in autonomous industrial vehicles, should be adapted to this use case. In this work, the focus has been set into the safety of the self-driving chassis, and the main goal has been to design a safe collision avoidance system. The core idea is to modify the current AGV (Automated Guided Vehicle) obstacle avoidance approach, which is based on warning fields, to be accommodated for the self-driving chassis paradigm. Furthermore, not only it has been pursued to readjust the state of the art solution, but a substantial improvement has also been aimed: to enable higher driving velocities while assuring the current safety levels are still achieved. To make faster velocities feasible, a novel safe collision concept has been introduced, with two prime innovative aspects. On the one hand, prediction has been incorporated through a neural network LSTM (Long Short Term Memory). A manner to integrate prediction uncertainty has been designed as well, to provide a robust behaviour. On the other hand, the standing pedestrian problem has been addressed. A simple but effective algorithm has been suggested to properly reduce the vehicle speed when driving next to a pedestrian that is standing. Such a pedestrian intention is hard to foresee, meaning it could start moving at any moment. The proposed new concept has been tested with exhaustive simulations. Then, special attention has been payed to the international standard ISO 13849, which is the state of the art functional safety standard for AGVs and control parts of machinery in general. The procedure defined in the standard has been followed to see the safety level that the system could reach. In order to calculate the safety level of a such complex system, which includes black-box algorithms like neural networks, a Markov chain has been designed. A node diagram has been presented which should serve as a good generalization of the possible states in a potential collision scenario. After estimating its parameters based on the simulation results, it has been proved that the current safety level, which is PL = d in ISO 13849 terms, is theoretically possible at a speed three times faster than the state of the art velocity. This work has concluded with the testing in real experiments, to validate the simulation assumptions and verify that a material implementation is conceivable
Collections
Files | Description | Size | Format | View |
---|---|---|---|---|
safe-collision- ... -mobile-assembly-final.pdf![]() | 3,474Mb | Restricted access |
All rights reserved. This work is protected by the corresponding intellectual and industrial
property rights. Without prejudice to any existing legal exemptions, reproduction, distribution, public
communication or transformation of this work are prohibited without permission of the copyright holder