Learning a Lattice Planner Control Set for Autonomous Vehicles

Abstract

This paper introduces a method to compute a sparse lattice planner control set that is suited to a particular task by learning from a representative dataset of vehicle paths. To do this, we use a scoring measure similar to the Fréchet distance and propose an algorithm for evaluating a given control set according to the scoring measure. Control actions are then selected from a dense control set according to an objective function that rewards improvements in matching the dataset while also encouraging sparsity. This method is evaluated across several experiments involving real and synthetic datasets, and it is shown to generate smaller control sets when compared to the previous state-of-the-art lattice control set computation technique, with these smaller control sets maintaining a high degree of manoeuvrability in the required task. This results in a planning time speedup of up to 4.31x when using the learned control set over the state-of-the-art computed control set. In addition, we show the learned control sets are better able to capture the driving style of the dataset in terms of path curvature.

Learning-based Fast Path Planning in Complex Environments

“…Zhang et al [18] implement a policy-based search method to learn an implicit sampling distribution for specific environments. De et al [19] propose to learn a lattice planner control set to achieve path planning for autonomous vehicles. The imitation learning methods such as conditional variational auto-encoder (CVAE) [20], generative adversarial network (GAN) [21] and recurrent neural network (RNN) [22] are used to bias the search direction through various probabilistic models.…”

Efficient Motion Planning for Automated Lane Change based on Imitation Learning and Mixed-Integer Optimization

“…Two main disadvantages of MPC were addressed: computation efficiency and requirement of full observation of the system. Reference [11] introduced an approach to learning a lattice planner control (LLPC) set, the controller was learned from a representative dataset of vehicle paths and shown to be suitable for one particular task. This model is able to imitate the driving style of the learning path and easy to compute.…”

A trajectory deformation algorithm for intelligent vehicles

“…Learning methods have been applied to improve lattice-based planners. Iaco, Smith and Czarnecki (2019) propose to learn from a representative data set of vehicle paths in order to compute a sparse lattice planner control set that is suited for a particular application. They develop an algorithm to evaluate a given control set according to a scoring measure.…”