A trajectory describes the path that the vehicle should follow within the next few seconds, as well as the speed profile that specifies a target speed for each point along the path. In generating such a trajectory, a number of partially contradictory goals must be taken into account: Among other things, it should bring the car swiftly closer to its destination, but at the same time avoid static obstacles (such as parked cars or trees) and dynamic obstacles (such as moving cars or pedestrians) and be comfortable to drive, i.e. do not require jerky steering or braking manoeuvres.
At the Institute for Autonomous Systems Engineering, trajectories are described using clothoids, which are also used in the planning of road courses due to their properties that promote driving comfort. A trajectory consists of several clothoid segments along which the vehicle constantly accelerates or brakes. Depending on the application, different methods are used to generate the optimal trajectory.
For autonomous navigation along waypoints (e.g. GPS coordinates), a variant of the Hybrid A* algorithm is used. Using a map created from sensor data, which divides the vehicle environment into square cells and provides information such as obstacle probability, gradient and road probability for each cell, it calculates the optimal path to the current destination. Then, taking into account the dynamic obstacles in the environment, the one that offers the best compromise between safety, driving time and driving comfort is selected from various possible speed profiles.
In the so-called convoy scenario, the vehicle should follow the path of a command vehicle tracked with the help of the on-board sensors as precisely as possible. In contrast to waypoint navigation, a procedure based on Sequential Quadratic Programming (SQP) is used here, which iteratively adjusts the parameters of an initial trajectory (the start solution) using a cost function to be minimized. The cost function includes properties such as longitudinal and lateral acceleration, the distance to the command vehicle and the deviation from the path of the command vehicle.
