No abstract

In automated parking systems, a path planner generates a path to reach the vacant parking space detected by a perception system. To generate a safe parking path, accurate detection performance is required. However, the perception system always includes perception uncertainty, such as detection errors due to sensor noise and imperfect algorithms. If the parking path planner generates the parking path under uncertainty, problems may arise that cause the vehicle to collide due to the automated parking system. To avoid these problems, it is a challenging problem to generate the parking path from the erroneous parking space. To solve this conundrum, it is important to estimate the perception uncertainty and adapt the detection error in the planning process. This paper proposes a robust parking path planning that combines an error-adaptive sampling of generating possible path candidates with a utility-based method of making an optimal decision under uncertainty. By integrating the sampling-based method and the utility-based method, the proposed algorithm continuously generates an adaptable path considering the detection errors. As a result, the proposed algorithm ensures that the vehicle is safely located in the true position and orientation of the parking space under perception uncertainty.

There are multifarious stationary vehicles in urban driving environments. Autonomous vehicles need to make appropriate overtaking maneuver decisions to navigate through the stationary vehicles. In literature, overtaking maneuver decision problems have been addressed in the perspective of either discretionary lane-change or parked vehicle classification. While the former approaches are prone to generating undesired overtaking maneuvers in urban traffic scenarios, the latter approaches induce deadlock situations behind a stationary vehicle which is not distinctly classified as a parked vehicle. To overcome the limitations, we analyzed the significant decision factors in the traffic scenes and designed a Deep Neural Network (DNN) model to make human-like overtaking maneuver decisions. The significant traffic-related and intention-related decision factors were harmoniously extracted in the traffic scene interpretation process and were utilized as the inputs of the model to generate overtaking maneuver decisions in the same manner with the human driver. The overall validation results convinced that the extracted decision factors contributed to increasing the learning performance of the model, and consequently, the proposed decision-making system enabled the autonomous vehicles to generate more human-like overtaking maneuver decisions in various urban traffic scenarios.

Car-following control is a fundamental application of autonomous driving. This control has multiple objectives, including tracking a safe distance to a preceding vehicle and enhancing driving comfort. Model Predictive Control (MPC) is a powerful method due to its intuitiveness and capability to cover multiple objectives. MPC determines the relative importance of objectives through a set of weight factors, depending on which, the controller's behavior changes even if the traffic situations are the same. However, determining the optimal weight is not a trivial problem because there is no benchmark to evaluate the performance of the weight, and searching for weight factors with repeated driving experiments is timeconsuming. To solve this problem, we proposed an automatic tuning method to determine the weights of the MPC based on personal driving data. Personal driving data under naturalistic driving conditions provide car-following situations and driver's behaviors. These data can generate a reference model to represent the driver's driving style. Based on this model, the proposed method defined the automatic tuning problem as an optimization problem that minimizes the difference between the reference and the controller's response using the optimal weight factors. This optimization problem was solved using the Particle Swarm Optimization algorithm. The proposed method was implemented with an embedded optimization coder in an offline fashion. Its performance was evaluated using personal driving data. From this, the proposed method can reduce the effort and time required for an engineer to find the optimal weight factors.