pyfollower 0.1.4

Follower Crowd Simulation

Follower Crowd Simulation

Follower Crowd Simulation (Follower) is a project that simulates the movement of a crowd/robots of comformists with self-organizing characteristic. In Follower, ORCA is adapted as the basic collision avoidance model. This project is writen in C++ and built using CMake. pyfollower is the python interfaces for Follower with many easy-to-use APIs.

Getting started

To get started with Follower Crowd Simulation, you'll need to have Python 3.7+ installed on your system. You can download Python from the official website, and Follower Crowd Simulation can be installed using pip:

pip install pyfollower

The installation process of this project requires Cmake and C++ compilation environment, so if you are a Windows user, it is a better choice to use a compiled wheel file

pip install pyfollower-xxx.whl

After you have installed it, you can simply use this simulation Engine. As an example(Please make sure the numpy and matplotlib is installed):

import numpy as np
from pyfollower import FollowerEngine

# Initial scenario
dest = dict()
N = 50
sim = FollowerEngine(agent_radius=0.5)
sim.pref_velocity_correction(1)

for idx, i in enumerate(np.linspace(0, 1, N + 1)):
    if i == 1: break
    theta = i * np.pi * 2
    ox, oy = np.cos(theta), np.sin(theta)
    dx, dy = np.cos(theta + np.pi), np.sin(theta + np.pi)
    t = np.array([ox, oy, dx, dy]) * 20
    agent_id = sim.add_agent(*t)
    dest[agent_id] = t[-2:]

# Test obstacles
sim.add_obstacles([(5, 5), (-5, 5), (-5, -5), (5, -5)])
sim.process_obstacles()
obs = np.array([(-5, -5), (-5, 5), (5, 5), (5, -5), (-5, -5)])

# Run simulation --- Main loop
traj = []
for i in range(100):
    if not i % 10: print(sim.time)
    x = sim.get_agent_positions()
    for agent_id in range(N):
        dx = dest[agent_id] - x[agent_id, :]
        dist = np.sqrt(np.sum(dx ** 2))
        if dist < 0.5:
            prev = (0, 0)
        else:
            prev = 1.0 * dx / dist
        sim.set_agent_pref(agent_id, *prev)
    traj.append(x)
    print(x.T)
    sim.follower_step()

# Plot
import pylab as pl
import matplotlib.cm as cm
fig = pl.figure('Trajectory', facecolor='thistle', figsize=(4, 4))
ax = fig.gca()

traj = np.stack(traj)

pl.xlim(-20, 20)
pl.ylim(-20, 20)
ax.imshow(np.zeros((40, 40), np.bool), extent=(-20, 20, -20, 20), cmap='viridis')\
    .format_cursor_data = lambda data: ""

colors = cm.hsv(np.linspace(0, 1, N))
for i in range(N):
    ax.plot(traj[:, i, 0], traj[:, i, 1], c=colors[i])
ax.scatter(*x.T, c=range(N), s=20, cmap='hsv')
ax.plot(obs[:, 0], obs[:, 1], c='white')

ax.format_coord = lambda x, y: f'({x:^6.1f}, {y:^6.1f})'
[spine.set_color('white') for spine in ax.spines.values()]
ax.tick_params(axis='both', colors='white')

pl.show()

Citation

During installation, due to differences in individual machines, various tricky issues may arise. If you encounter any difficulties using this simulator, please do not hesitate to send an email to [e-mail].

If you find our work helpful (or if you are so kind as to offer us some encouragement), please consider citing the paper.

@misc{liao2024emergent,
      title={Emergent Crowd Grouping via Heuristic Self-Organization}, 
      author={Xiao-Cheng Liao and Wei-Neng Chen and Xiang-Ling Chen and Yi Mei},
      year={2024},
      eprint={2407.00674},
      archivePrefix={arXiv},
      primaryClass={cs.MA},
      url={https://arxiv.org/abs/2407.00674}, 
}