forrest-z / crowdnav

Robot Crowd Navigation through Predictive Pedestrian Modeling

The control stack for this robot has been built based on this repo. The obstacle detector package is provided by this repo and the pedestrian simulator is provided by this repo.

We present a local path planning algorithm, deployable on a service robot, that generates probabilistic models about the future as outlined in our paper in paper/. The implementation can be found in control_stack/include/cs/path_finding/rrt. This algorithm can be deployed on a service robot in different simulation environments and demonstration videos can be found in demos/.

Requirements

• ROS Melodic

Installation

• Clone this repo to your <catkin workspace folder>/src

Usage

• Ensure that you are in the root directory of your catkin workspace

$ source ./devel/setup.bash

• To run a single obstacle environment:

$ roslaunch pedsim_simulator 700_1obst.launch

• To run a hallway with pedestrian traffic against the service robot:

$ roslaunch pedsim_simulator 700_against_1way.launch

• To run a hallway with two way pedestrian traffic:

$ roslaunch pedsim_simulator 700_2way_hallway.launch

• To run a four way interection:

$ roslaunch pedsim_simulator 700_4way.launch

• Note that for each scenerio you want to run, you will need to update the map path in the corresponding nav_config_700_*.lua file in control_stack/config to match your absolute path

• Analytics for each run will be automatically printed on the terminal

Contributors

• Sebastian Peralta
• Raymond Bjorkman
• Zachary Fischer