Collision Avoidance in Tightly-Constrained Environments without Coordination: a Hierarchical Control Approach.
Citations
5 citations
3 citations
1 citations
1 citations
1 citations
References
1,539 citations
"Collision Avoidance in Tightly-Cons..." refers methods in this paper
...These were collected using the CARLA simulator [18] in a custom parking lot [19], where the subject (driver) controls the brake, throttle, and steering of the vehicle via a Logitech G27 racing wheel and pedals (Fig....
[...]
1,538 citations
"Collision Avoidance in Tightly-Cons..." refers background or methods in this paper
...The problem of collision avoidance is nonconvex and NP-hard in general [10]....
[...]
...Both BL and SG approaches were implemented with the same cost matrices Qz = diag([1, 1, 1, 10]), Qu = diag([1, 1]), and Qd = diag([50, 50]), on the 1/10 scale open source Berkeley Autonomous Race Car (BARC) platform in a laboratory parking lot environment (Fig....
[...]
1,437 citations
1,162 citations
"Collision Avoidance in Tightly-Cons..." refers background in this paper
...Hierarchical approaches for autonomous driving have therefore been proposed in the literature, relying on path planning at the top level, maneuver choices at the strategic level, and trajectory planning and control at the lowest level [1], [5], [6], [8], [9]....
[...]
1,142 citations
"Collision Avoidance in Tightly-Cons..." refers methods in this paper
...One interpretation of the strategy-guided constraints is that they are surrogates of the MPC terminal components, where given the selected strategy, we attempt to construct constraints which drive the state of the system (2) into a set which is contractive towards the reference zref while satisfying all constraints....
[...]
...For the lower-level online strategy-guided control scheme, we construct three policies: ΠSG = {πSG−OBCA, πSC, πEB}, (5) where πSG−OBCA is a nominal MPC controller based on OBCA [12], [17], which uses the selected strategy to generate hyperplane constraints which help reduce myopic behavior and improve the ability of the EV to complete its navigation task. πSC is a safety controller which can be activated due to infeasibility of SG-OBCA or ambiguity in the behavior of the TV. πEB is an emergency brake controller, to be triggered in the event of impending collision....
[...]
...In contrast, the baseline control scheme only reverts to the safety control when the nominal MPC becomes infeasible....
[...]
...The MPC predictions at each time step are used to construct OTVk over the EV control horizon....
[...]
...For the nominal MPC policy πSG−OBCA in ΠSG, we leverage the generated constraints to extend the OBCA algorithm [12] which formulates the collision avoidance constraints using exact vehicle geometry to enable tight maneuvers in narrow spaces....
[...]