top of page

Long Term Multi-Destination
Autonomous Navigation Without HD Maps

Jinze Liu*, Dongmyeong Lee*, Rui Chen, Qi Dai, Dianhao Chen, Maani Ghaffari, Jiunn-Kai Huang and Jessy W. Grizzle
(* co-first author)

Abstract

We propose and experimentally demonstrate an autonomous navigation system for bipedal robots outdoor and indoor withough HD maps. The system consists of 1) a global route planning thread to define a route on a map for navigation, 2) a localization thread to estimate a pose of the robot on the map, and 3) a reactive thread to find a safe local path with a segmented walkable region surrounding the robot. The global route planning has two options; 1) Informable Multi-Objective and Multi-Destinations RRT* (IMOMD-RRT*) system that finds an asymptotically shortest path that visits all destinations and 2) Picture to OSM that captures user-drawn path on the printed map. The localization threads consists of two parts; 1) the front-end part is Feature Extraction that detects curb for outdoor experiments and wall for indoor experiements using LiDAR and 2) the back-end part is Localization that estimate pose of the robot using particle filter. The extracted featuers and landmarks of the map are lifted to a continuous function in a Reproducing Kernel Hilbert Space (RKHS) where the inner product can be used to determine weights for particles. The reactive planning consists of 1) terrain classification part and 2) CLF local planning part. First, the terrain classification part segments walkable regions surrounding the robot using Support Vector Machine (SVM) and dynamica programming with point clouds that projected onto the image. Second, the CLF local planning part select local goal based on the segmented local map and elevation map, then the CLF reactive planner generates an optimal control commands. The system is evaluated on various challenging outdoor terrains and indoor scenes on Cassie Blue, a bipedal robot with 20 degrees of freedom.

Experiment Video (Outdoor)

Overall Pipeline for the Autonomous Navigation System

Curb Feature Extraction Video (Outdoor)

Wall Feature Extraction Video (Indoor)

bottom of page