Section: New Results

Motion planning techniques

Participants : David Gonzalez Bautista, Fernando Garrido Carpio, Joshué Pérez Rastelli, Vicente Milanés Montero, Fawzi Nashashibi.

Intelligent vehicles have increased their capabilities for highly, and even fully, automated driving under controlled environments. Scene information is received using on-board sensors and communication network systems–i.e. infrastructure and other vehicles. Considering the available information, different motion planning techniques have been implemented to autonomously driving on complex environments. The main goal is focused on executing strategies to improve safety, comfort and energy optimization. However, research challenges such as navigation in urban dynamic environments with obstacle avoidance capabilities–i.e. Vulnerable Road Users (VRU) and vehicles–and cooperative maneuvers among automated and semi-automated vehicles still need further efforts for a real environment implementation. We have recently carried out a deep state-of-the-art review to find the gaps in this hot topic into the autonomous vehicle field, paying special attention to overtaking and obstacle avoidance maneuvers.

Based on this review, we have mainly identified two main gaps: trajectory and speed planning with dynamics obstacle avoidance capabilities and real-time performance of the algorithms in the sense of significantly reducing the computational time, moving the system closer to what a vehicle should be able to provide in the real world.

According to this review, a speed planner has been designed with specific considerations on computing time efficiency, with an optimal comfort and avoiding to exceed speed and acceleration limits [31] . The comfort is evaluated as the minimization and smoothness of acceleration and jerk profiles, while maintaining a coherent speed profile with respect to traffic rules, the geometry of the path and the lateral accelerations associated to it. Specifically, this speed planner uses fifth order polynomial curves. These curves are C2 continuous and smooth, meaning that the jerk profile is also continuous and smooth. The method proposed computes the velocity in terms of the length of the path, instead of time, greatly reducing the errors. Specific targets for the speed planner are:

• Compute a smooth and continuous speed profile accounting for acceleration limits (longitudinal and lateral) according to ISO 2631-1 standard.

• Minimize distance error problems by associating the speed profile in the path speed planner instead of the time.

This speed planner was tested against other techniques providing better results in terms of computational time and smoothness (cf. [32] ).

Additionally, a novel trajectory planning with a significant reduction on the computational time with respect to prior implementations from the team has been implemented. Our approach is mainly affected by vehicle's kinematics and physical road constraints. Based on these assumptions, computational time for path planning can be significantly reduced by creating a database containing already optimized versions of all the potential trajectories in each curve the vehicle can carry out. Therefore, this algorithm generates a database of smooth and continuous curves considering a big set of different intersection scenarios, taking into account the constraints of the infrastructure and the physical limitations of the vehicle. According to the real scenario, the local planner selects from the database the appropriate curves, searching for the ones that fit with the intersections defined on it. The path planning algorithm has been tested in simulation against the previous control architecture. The results obtained show path generation improvements in terms of smoothness and to continuity. Next steps on this algorithm is to test its performance in real platform and add the dynamics obstacle avoidance capabilities, establishing the link with the perception algorithms research line currently open in the team.