Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluating Robot Influence on Pedestrian Behavior Models for Crowd Simulation and Benchmarking

Subham Agrawal, Nils Dengler, Maren Bennewitz

TL;DR

A simulation framework that repetitively measures and benchmarks the deviation in trajectory of pedestrians due to robots driven by different navigation algorithms, resulting in the Social Robot Force Model (SRFM).

Abstract

The presence of robots amongst pedestrians affects them causing deviation to their trajectories. Existing methods suffer from the limitation of not being able to objectively measure this deviation in unseen cases. In order to solve this issue, we introduce a simulation framework that repetitively measures and benchmarks the deviation in trajectory of pedestrians due to robots driven by different navigation algorithms. We simulate the deviation behavior of the pedestrians using an enhanced Social Force Model (SFM) with a robot force component that accounts for the influence of robots on pedestrian behavior, resulting in the Social Robot Force Model (SRFM). Parameters for this model are learned using the pedestrian trajectories from the JRDB dataset. Pedestrians are then simulated using the SRFM with and without the robot force component to objectively measure the deviation to their trajectory caused by the robot in 5 different scenarios. Our work in this paper is a proof of concept that shows objectively measuring the pedestrian reaction to robot is possible. We use our simulation to train two different RL policies and evaluate them against traditional navigation models.

Evaluating Robot Influence on Pedestrian Behavior Models for Crowd Simulation and Benchmarking

TL;DR

A simulation framework that repetitively measures and benchmarks the deviation in trajectory of pedestrians due to robots driven by different navigation algorithms, resulting in the Social Robot Force Model (SRFM).

Abstract

The presence of robots amongst pedestrians affects them causing deviation to their trajectories. Existing methods suffer from the limitation of not being able to objectively measure this deviation in unseen cases. In order to solve this issue, we introduce a simulation framework that repetitively measures and benchmarks the deviation in trajectory of pedestrians due to robots driven by different navigation algorithms. We simulate the deviation behavior of the pedestrians using an enhanced Social Force Model (SFM) with a robot force component that accounts for the influence of robots on pedestrian behavior, resulting in the Social Robot Force Model (SRFM). Parameters for this model are learned using the pedestrian trajectories from the JRDB dataset. Pedestrians are then simulated using the SRFM with and without the robot force component to objectively measure the deviation to their trajectory caused by the robot in 5 different scenarios. Our work in this paper is a proof of concept that shows objectively measuring the pedestrian reaction to robot is possible. We use our simulation to train two different RL policies and evaluate them against traditional navigation models.
Paper Structure (21 sections, 7 equations, 3 figures, 3 tables)

This paper contains 21 sections, 7 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Our Approach
  4. Social Robot Force
  5. Learning SFRM parameters
  6. Deviation Metric and Simulation Scenarios
  7. Training
  8. Evaluation
  9. Trajectory deviation metric
  10. Other metrics
  11. Learning a Navigation Policy
  12. Actions
  13. Observations
  14. Reward
  15. Training method
  16. ...and 6 more sections

Figures (3)

  • Figure 1: Example scenario of a human navigating to a goal. While many of the forces used in the social force model are well studied, such as attraction to the goal (green arrow) or repulsion from obstacles, such as the pillar (blue arrow), the robot force (red arrow) is not well understood, even though it significantly affects human behavior.
  • Figure 2: Architecture of the simulation system. The Social Robot Force Model (SRFM) parameters are learned from the dataset. The learned SRFM is then used to drive the pedestrians in simulation while an RL agent drives the robot during training and evaluation phases.
  • Figure 3: Overview of the different scenarios used for the evaluation of navigation policies. (a) Training environment that is used for the RL agent. (b), (c), and (d) visualize the three common pedestrian streams as described in francis2023principles. (e) and (f) visualize two extreme scenarios.