Lattice-based Motion Planning for a General 2-trailer system

TL;DR

This paper introduces a lattice-based motion planning framework for a complex 2-trailer system that guarantees completeness and optimality, enabling real-time, obstacle-aware maneuvers in both forward and backward directions.

Contribution

It presents a novel resolution complete and optimal lattice-based planning method with offline precomputation and a new parametrization for real-time applicability.

Findings

Achieves real-time performance in complex scenarios

Guarantees resolution completeness and optimality

Handles both forward and backward maneuvers

Abstract

Motion planning for a general 2-trailer system poses a hard problem for any motion planning algorithm and previous methods have lacked any completeness or optimality guarantees. In this work we present a lattice-based motion planning framework for a general 2-trailer system that is resolution complete and resolution optimal. The solution will satisfy both differential and obstacle imposed constraints and is intended as a driver support system to automatically plan complicated maneuvers in backward and forward motion. The proposed framework relies on a precomputing step that is performed offline to generate a finite set of kinematically feasible motion primitives. These motion primitives are then used to create a regular state lattice that can be searched for a solution using standard graph-search algorithms. To make this graph-search problem tractable for real-time applications a novel parametrization of the reachable state space is proposed where each motion primitive moves the system from and to a selected set of circular equilibrium configurations. The approach is evaluated over three different scenarios and impressive real-time performance is achieved.