Simulation of Multiple, Wheeled Agents using Local Navigation

1
Simulation of Multiple, Wheeled Agents using
Local Navigation

• By David Wilkie, Tim Johnson, Yu-Hsien Su

2
Introduction

• We plan on adding car-like motion constraints to
  the Reciprocal Velocity Obstacle (RVO)
  formulation.
• Can simulate large numbers of agents
• Currently in RVO, agents can move freely in 2D.
• However, wheeled agents are subject to kinematic
  constraints and minimum turning radius.
• We will apply RVO with these new constraints to
  the problem of traffic simulation.

3
Introduction (cont)

• Many previous simulations of traffic relied on
  constraining the motion of cars to predefined
  lanes.
• These simulations cannot cope with unruly
  intersections or unanticipated agent actions.
• We propose to apply an agent-based formulation to
  these scenarios.
• Other simulation work in traffic does not take
  into account the reciprocal behavior in cars.
• We account for this by building off of the
  reciprocal velocity obstacle formulation. (Berg,
  Lin, Manocha)?

4
Motivation

• Create visually realistic simulations of noisy,
  real world traffic.

5
Motivation (cont)

• Model accurately the movement capabilities of
  wheeled robots within a multi-agent local
  navigation framework.

6
Motivation (cont)

• Create a simulation framework that could be
  applied to traffic simulations with heterogeneous
  agents moving without strict constraints of lanes
  and traffic rules.

7
Prior Art

• General Traffic Simulation
• Microscopic Traffic Simulation
• Follow-the-leader model (Reuschel Pipes)
• Cellular automata (Nagel Schadschneider)
• Master equation
• Macroscopic Traffic Simulation
• Lighthill and Whitham model (Lighthill Whitham)

8
Prior Art (cont)

• Traffic Simulation Focusing on Intersections
• Cellular Automata (Jin, Zheng, Li)
• Navigation Algorithms
• Velocity Obstacles (VO) (Fiorini, Shiller)
• Reciprocal Velocity Obstacles (RVO) (Berg, Lin,
  Manocha)
• Nonlinear Reciprocal Velocity Obstacles (NRVO)

9
Goal of Project

• Traffic simulator
• Focus on busy intersections
• Agent-based
• Integrate decision model for cars
• Complete freedom (subject to constraints) in 2D
  plane
• http//www.youtube.com/watch?vPySM8dGaNLo

10
Novelty

• Integration of non-holonomic constraints into RVO
• Agent-based inference of reciprocity
• How much effort will agent B make to avoid agent
  A?
• Model roads as fully-continuous, 2 dimensional