Abstract
Automated vehicles will have to achieve the level of performance of experienced human drivers in order to satisfy passenger’s expectations and be able to share the road with manually driven cars. This imposes strict requirements for automated driving systems and motion planning in particular. Although robot motion planning has been intensively studied in the last decades, questions remain in the context of automated driving. A planning solution has to ensure that the whole behavioral instruction set with its varying complexity can be fully expressed. However, a separation of the planning process into several less complex solutions is often the cause for complex behaviors not being adequately expressed. The objective of this study is to analyze the state-of-the-art in motion planning and identify methods that allow the modeling of such behavioral situations. General benchmark criteria for these methods have been defined. Existing solutions have not been proven to be adequate in all respects. Therefore, in this paper a newly developed high dimensional phase space (covering position, velocity, and momentum) approach is proposed. The described search space unifies all available environment information, including object dynamics to generate behaviors. Therefore, objects must not directly interact with the vehicle in order to be treated as valuable planning input.
KEYWORDS - motion planning, sampling-based planning, behavior adaption, complex vehicle maneuvering