Dynamic Trajectory Planning for Autonomous Driving Based on Fluid Simulation

Autonomous Driving Using Fluid Flow Simulation

“…Once a sufficient number of cycles of fluid simulation is completed, a path is planned using velocity vectors in each lattice cell. A variation in this approach, where the source of the fluid is positioned behind the controlled vehicle, has proven to be a reliable method for dynamically driving a car [20].…”

“…A reliable trajectory must be planned to drive a vehicle autonomously. This can be achieved by simulating fluid flow in a river system [22] that may be derived from a road layout [20]. Areas of higher and lower pressure are required to achieve fluid flow.…”

“…Its cells contained friction values that affected how fast the simulated molecules propagated between the grid cells. As per [20], the LBM pressure source was placed at a 45degree angle, 10 meters behind the driven car, to generate a fluid flow vector map pointing forward and slightly to the right side. To encourage forward motion, each grid border except for the back one (relative to the driven car) has constant minimum pressure.…”

“…When compared to lane-basing methods, CFF does not commit to driving on lanes, as it does not possess information about them. Despite this disadvantage, the resulting trajectories prove to be feasible [20] due to the balance of fluid source angle, fluid propagation speed and LBM cell size. This feature should be very useful when driving over a road without marked lanes present.…”

“…Trajectory planning is performed by simulating fluid flow in a river system based on a road network. A feasible trajectory can be found using this approach, as fluid flows along the path of least resistance [20]. Additional steps are taken to ensure that the trajectory is smoothed according to vehicle physics.…”