human-robot-collider

This software package simulates and analyses collisions between a walking pedestrian model and a mobile robot to investigate the risk associated to such human-robot collisions and to support safety-guided robot design. The software lets both agents approach each other and collide with different angles and speeds, using the simulation engine Bullet (via pybullet).

Prerequisites

It is recommended to use a virtual environment to setup the simulation. Process to setup the environment and install all dependencies is given below.

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Executing simulation

The simulation can be excuted by the following command,

python path/to/human-robot-collider/src/hrc.py

Argument	Options	Help
-h, --help		Show the help message and exit
-b, --human	man, child	Human to collide with the robot (default = man)
-r, --robot	qolo, wheelchair, pepper	Robot to collide with the human (default = qolo)
-c, --controller	no_control, admittance, passive_ds	Adaptive controller to use (default = no_control)
-g, --gui		Set to show GUI

If --gui argument is activated, plot of collision force are drawn after each collision.

At the end of all simulations, collision_forces array for all collisions are saven in controlled_collision.npy for future use.

Examples

• To run simulation of collision between a man and qolo using admittance controller (with gui),

python src/hrc.py -b adult -r qolo -c admittance -g

• To run simulation of collision between a child and qolo using admittance controller (with gui),

python src/hrc.py -b child -r qolo -c passive_ds -g

• To run simulation of collision between a child and qolo using no compliance controller (with gui),

python src/hrc.py -b child -r qolo -c no_control -g