This thesis aims to enable efficient trajectory planning for autonomous vehicles without the requirement of a map or prior knowledge of the environment. For this purpose, an approach is presented, which adapts the concept of trail pheromones used by ants as well as the collective animals' flocking behavior to the domain of autonomous vehicles. In this way, the drivers of surrounding vehicles are used as additional input for anticipatory driving, expanding the capabilities of the individual autonomous car: a typical achievement of swarm intelligence.
While map-based trajectory planning has many advantages, there are situations when no map is available or localization in a map is too inaccurate. Also, the actual behavior of road users may differ significantly from the given map. Existing approaches for planning driving maneuvers in urban traffic without a map decouple lateral and longitudinal planning. Usually, knowledge of road geometry is assumed. This thesis presents an approach to overcome those limitations, adopting the established theory of elastic bands to implement swarm-based trajectory planning. The vehicle's dynamic restrictions and the driver's preferences are represented with a comprehensive set of parameterized objective functions. A swarm-based motion prediction algorithm is introduced to predict the surrounding vehicles' trajectories, and a heuristic is presented to choose the best candidate for a leader vehicle based on weighted criteria. The approach is evaluated in simulation, on recorded data, and live field tests in real traffic. The experimental results show that the presented approach, implementing a swarm behavior for autonomous cars, is valid to temporarily compensate for the advantages of map-based planning.