Automated driving: The role of forecasts and uncertainty—A control perspective

Abstract

Driving requires forecasts. Forecasted movements of objects in the driving scene are uncertain. Inevitably, decision and control algorithms for autonomous driving need to cope with such uncertain forecasts. In assisted driving, the uncertainty in the human/vehicle interaction further increases the complexity of the control design task.Our research over the past ten years has focused on control design methods which systematically handle uncertain forecasts for autonomous and semi-autonomous vehicles. This article presents an overview of our findings and discusses relevant aspects of our recent results.

A game theoretical model of traffic with multiple interacting drivers for use in autonomous vehicle development

“…They are not necessarily optimal driver profiles or representative of most drivers. One area for future work is the characterization of actual driver profiles, and the primary use of these profiles is to illustrate the usefulness of the proposed methods for Figure 6 presents the results from a series of simulations, where each curve shows the average over 1000 trials for each n c ∈ [1,60], which is the abscissa in the figures. The simulations have 3 lanes, and the duration is set to t f = 200sec with x 0 max = 600m and the maximum visual range set to 400m.…”

“…As is apparent from the prior literature (see e.g., [1] [2] [3]), models of human driver actions in a given traffic scenario can be exploited for the development of decision making algorithms for autonomous driving and for the implementation of high-fidelity simulators that can facilitate the validation and testing of competing autonomous driving policies. A comprehensive list of existing human driver models, control based and behavioral based, can be found in [4].…”

Hierarchical reasoning game theory based approach for evaluation and testing of autonomous vehicle control systems

“…Such simulators can save time in the development phase by providing a model-based testing environment, before the actual road tests. Secondly, these models can be used in the design of hierarchical control schemes for driverless cars: typically, in an autonomous vehicle, a higher level outer loop controller generates the reference trajectories for the lower level inner loop controller, which determines the steering angles, acceleration/deceleration inputs, etc., required to follow the reference trajectory [1]. Predictive driver models can be utilized in the higher level outer loop controller generating the reference trajectories for the lower level inner loop controller, thereby ensuring similar behavior to that of a human-driven vehicle and improving the comfort level of the passengers [1].…”

“…Secondly, these models can be used in the design of hierarchical control schemes for driverless cars: typically, in an autonomous vehicle, a higher level outer loop controller generates the reference trajectories for the lower level inner loop controller, which determines the steering angles, acceleration/deceleration inputs, etc., required to follow the reference trajectory [1]. Predictive driver models can be utilized in the higher level outer loop controller generating the reference trajectories for the lower level inner loop controller, thereby ensuring similar behavior to that of a human-driven vehicle and improving the comfort level of the passengers [1]. In addition, these models can provide predictions of the future trajectories of the vehicles in the vicinity of the host autonomous vehicle and be used as inputs for the inner loop controllers such as model predictive controllers (MPC) [2]- [4].…”

Game Theoretic Modeling of Driver and Vehicle Interactions for Verification and Validation of Autonomous Vehicle Control Systems

“…1), wherein a higher level outer-loop controller generates reference trajectories for the lower level inner-loop controller, which, in turn, determines the steering angle and acceleration/deceleration inputs required to track the reference trajectory [8].…”