Problem

Autonomous vehicles must be able to react quickly and reliably in safety-critical situations. In today’s modular software architectures, however, perception and trajectory planning often operate largely separately: perception provides only simplified object information, while uncertainties regarding object position, existence, or classification are only used to a limited extent in planning. As a result, chains of errors can occur, limiting the safety and performance of the overall system, especially during highly dynamic evasive maneuvers.

Objective

The objective of the PiP2 project is to develop new software modules for autonomous vehicles that more closely integrate perception and trajectory planning. Through bidirectional information exchange, planning information will be used to detect relevant objects more selectively, while uncertainties from perception will be directly incorporated into trajectory planning. The developed approaches will first be tested in simulation, then on a HiL test bench, and finally in a real vehicle, enabling safe and agile driving maneuvers in critical situations.