Real-time Trajectory Planning to Enable Safe and Performant Automated Vehicles Operating in Unknown Dynamic Environments

Abstract

Need for increased automated vehicle safety and performance will exist until control systems can fully exploit the vehicle's maneuvering capacity to avoid collisions with both static and moving obstacles in unknown environments. A safe and performance-based trajectory planning algorithm exists that can operate an automated vehicle in unknown static environments. However, this algorithm cannot be used safely in unknown dynamic environments; furthermore, it is not real-time. Accordingly, this thesis addresses two overarching research questions:

* How should a trajectory planning algorithm be formulated to enable automated ground vehicle safety and performance in unknown dynamic environments?

* How can such an algorithm be solved in real-time?

Safe trajectory planning for high-performance automated vehicles with both static and moving obstacles is a challenging problem. Part of the challenge is developing a formulation that can be solved in real-time while including the following set of specifications: minimum time to goal, a dynamic vehicle model, minimum control effort, both static and moving obstacle avoidance, simultaneous optimization of speed and steering, and a short execution horizon. This thesis presents a nonlinear model predictive control-based trajectory planning formulation, tailored for a high mobility multipurpose wheeled vehicle (HMMWV), that includes the above set of specifications. This formulation is tested then with various sets of these specifications in a known dynamic environment. In particular, a parametric study relating execution horizon and obstacle speed reveals that the moving obstacle avoidance specification is not needed for safety when the planner has a short execution horizon (< 0.375 s), and the obstacles are slow (< 2.11 m/s). However, a moving obstacle avoidance specification is needed when the obstacles move faster, and this specification improves safety without, in most cases, increasing solve-times. Overall, results indicate that trajectory planners for high-performance automated vehicles should include the entire set of specifications mentioned above unless a static or low-speed environment permits a less comprehensive planner.

Then, this thesis combines this comprehensive planning algorithm with a suitable perception algorithm to enable safe and performant control of automated ground vehicles in unknown dynamic environments. A high-fidelity, ROS-based proving ground with a 2D LiDAR model, in Gazebo, and a 145 degree of freedom model of the HMMWV, in Chrono, is developed to combine these algorithms. Six-hundred tests, realized with various obstacle speeds and sizes, are performed in this proving ground in both known and unknown dynamic environments. Results from this comparison demonstrate that operating in an unknown environment, as opposed to a known environment, significantly increases collisions, steering effort, throttle effort, braking effort, orientation and tracking error, time to goal, and planner solve times. To avoid this deterioration of safety and performance factors in unknown environments, the use of more accurate perception systems should be explored. Ultimately, however, these results demonstrate that the comprehensive trajectory planning formulation developed in this thesis enables safe and performant control of automated vehicles in unknown dynamic environments among small (< 2 m) obstacles traveling at speeds up to high (20 m/s).

To solve this formulation in real-time, an open-source, direct-collocation-based optimal control problem modeling language, called NLOptControl, is established in this thesis. Results demonstrate that NLOptControl can solve the formulation in real-time in both known and unknown environments. NLOptControl holds great potential for not only improving existing off-line and on-line control systems but also engendering a wide variety of new ones.