Jointly Learnable Behavior and Trajectory Planning for Self-Driving Vehicles

Abstract

The motion planners used in self-driving vehicles need to generate trajectories that are safe, comfortable, and obey the traffic rules. This is usually achieved by two modules: behavior planner, which handles high-level decisions and produces a coarse trajectory, and trajectory planner that generates a smooth, feasible trajectory for the duration of the planning horizon. These planners, however, are typically developed separately, and changes in the behavior planner might affect the trajectory planner in unexpected ways. Furthermore, the final trajectory outputted by the trajectory planner might differ significantly from the one generated by the behavior planner, as they do not share the same objective. In this paper, we propose a jointly learnable behavior and trajectory planner. Unlike most existing learnable motion planners that address either only behavior planning, or use an uninterpretable neural network to represent the entire logic from sensors to driving commands, our approach features an interpretable cost function on top of perception, prediction and vehicle dynamics, and a joint learning algorithm that learns a shared cost function employed by our behavior and trajectory components. Experiments on real-world self-driving data demonstrate that jointly learned planner performs significantly better in terms of both similarity to human driving and other safety metrics, compared to baselines that do not adopt joint behavior and trajectory learning.

MP3: A Unified Model to Map, Perceive, Predict and Plan

“…Motion forecasting of dynamic objects becomes even more challenging without having access to the lanes that vehicles typically follow or the location of crosswalks for pedestrians. Most importantly, the search space to plan a safe maneuver for the SDV goes from narrow envelopes around the lane center lines [1,45,46,50] to the full set of dynamically feasible trajectories as depicted in Fig. 1 (right).…”

Deep Structured Reactive Planning

“…More recently, learning-based cost functions also show promising results. Those costs can be learned through either Imitation Learning [44] or Inverse Reinforcement Learning [61]. In most of these systems, predictions are made independently of planning.…”

“…We define the collision energy to be γ if a pair of future trajectories collide and 0 if not. Following [44], we define the safety distance violation to be a squared penalty within some safety distance of each actor's bounding box, scaled by the speed of the SDV. In our setting, we define safety distance to be 4 meters from other vehicles.…”

Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior

“…AV Planners. Despite the recent academic interest in endto-end learning-based planners and AVs [3,7,44,45,58], rule-based planners remain the norm in practical AV systems [52]. Therefore, we evaluate our approach on a rule-based planner similar to the lane-graph-based planners used by contestants in the 2007 DARPA Urban Challenge [36,50] detailed in Sec.…”