Section: New Results

Motion planning techniques

Participants : Fernando Garrido, David González Bautista, Fawzi Nashashibi.

Overtaking and lane change maneuvers represent some of the major causes of fatalities in road transport. The role of the path planning in these maneuvers is essential, not only for designing collision-free trajectories, but also to provide comfort to the occupants of the vehicle.

Having this in mind, a novel two-phase dynamic local planning algorithm to deal with these dynamic scenarios has been proposed, based on previous work. In the first phase (pre-planning) [47], a multi-objective trajectory optimization considering static information (i.e. digital maps) is carried out, using quartic Bézier curves as the path generation, which let us consider the constraints of both vehicle and road, generating continuous paths in the next phase. In the second phase (real-time planning) [46], time-horizon based trajectory generation is provided on a real-time using the pre-planned information. A human-like driving style is provided evaluating the sharpness of the road bends and the available space among them, smoothing the path. There, the paths are generated by joining the already optimized quartic Bézier curves ensuring continuity in the transitions among bends and straights.

Based on this architecture, a dynamic path planning approach has been introduced to safely avoid the possible obstacles in the path. A grid based solution has been developed to discretize the space and process the obstacles. It computes a virtual lane that re-plans the local path to be tracked by modifying the global itinerary using a geometric approach considering dynamics of both overtaking and overtaken vehicles to find smooth lane changes. That way, the dynamic problem can be addressed with the described real-time static local planner. Then, the overtaking path is built by joining two curves for each lane change, minimizing the slopes, according to the virtual lane configuration, loading these curves from the pre-planning stage.

The proposed architecture has been validated both on simulation (with Pro-Sivic and RTMaps) and on the Inria Rocquencourt terrain (with Cybercars and a Citroen C1) for the static scenario, and on simulation for the dynamic scenario. The results showed a smoother tracking of the curves, reduction on the execution times and reduced global accelerations increasing comfort. Future works will improve the capacity to deal with unexpected circumstances while making the overtaking maneuvers, testing with different car types as obstacles.