B.Tech. Project Stage

B.Tech. Project Presentation
Intercepting a Moving Target in
Road Networks
by
Prateek Khatri
Under the guidance of
Prof. N. L. Sarda
The Problem


Given a road-network, n number of pursuers and one
evader, devise a strategy to coordinate all pursuers to
capture the evader.
Assumptions:



Speed of pursuer and evader not bounded
Pursuer receiving regular updates about evader position
Pursuer knows the initial position of the evader
Introduction

Devising the strategy: Here the aim is to develop a strategy
for pursuers considering the constraints of a road network.
Presently a simple shortest path strategy is implemented.
Developing an intelligent strategy for evader can help check
the efficiency of the pursuer strategy.

Simulation: Here the aim is to develop a web based
interface for simulation and analysis of the strategies.
The interface will allow the user to select the simulation
parameters like starting nodes for pursuer and evader,
no. of pursuers and no. of evaders, etc.
Previous Works





The work by Parsons and Motwani have focussed on the
visibility based pursuit-evasion in graphs
Some others have advocated the use of randomized solutions
as the probability of pursuer catching evader increases
Most of the works have focussed on polygonal environments
None of the work encountered have focussed on road
networks specifically
Randomized strategy as given in [2] using RRTs focusses on
polygonal regions but can be adapted for graphs as well
Limitations of earlier works



Road networks are very different from the robotic
environments.
Dynamic constraints on fuel, roads,traffic conditions, number
of vehicles available
Implicit assumptions:
 Bounded and polygonal environment
 No constraints on paths
 No constraints on number of pursuers
Strategies
Possible pursuer strategies:




Shortest path to evader at every update (Implemented)

Dividing the area into n parts for n pursuers

Randomized Strategy

Heuristic based strategies: roadblocks, toll booths, etc.
Possible evader strategies:

Random (Implemented)

Moving away from the initial point

Heuristic based strategies: crowded roads, narrow roads, hiding place, etc
Capture Conditions:

Pursuer within some small distance of evader (Implemented)

Pursuer can see evader (in case of line of sight)
Simulation
Discrete-event simulation has been implemented to test and
analyse the strategies.
The problem is simulated with one pursuer and one evader with
following strategies:






Pursuer – Shortest path at every update
Evader – Random run and moving far away from the initial position
Capture condition – evader within some distance of pursuer
Assumptions



Pursuer needs random updates to follow evader
Total number of events in the simulation can not be more than 1000
Simulation is over if it one of the two conditions are satisfied:


Evader is caught
Total number of events become more than 1000
Implementing Simulation
Implementing Simulation
Visualization
 A web-based visualization software is
developed to monitor and analyze the process
 User can set the simulation parameters, can
select the initial nodes for pursuer and evader.
 Developed using JSP, Servlets and OpenLayers
DEMO
Visualization workflow
index.jsp
controller.java
map.jsp
• user selects the map
• sets the no. of pursuers and
evaders
• map id is passed to the
controller class
• fetches the map data using the
map id.
• prepares a mapInfo obj
• stores the obj in the session
• user selects the initial nodes
• sets the simulation parameters
• run the simulation
showSimResult.jsp
controller.java
• user monitors the simulation
process
• user can control the process by
advancing the simulation
• prepares a Simulate obj.
• sets the simulation params
• run the simulation
• stores the simObj in the session
Manual Tasks
Extracting
Nodes
• Nodes (joints in a MULTILINESTRING where LINESTRINGS meet), needs to be extracted using
some external software, e.g.: QuantumGIS or a specific program written for this
• Convert the map shapefile to map<id>.xml
SHP to GML • Convert the map-nodes shapefile to map<id>nodes.xml
conversion
PostGIS
tables
• Create a table map<id> with following columns:
• gid – road ids in the map
• lines – geometry column containing roads as MULTILINESTRING geometry
• Create a table map<id>_nodes with following columns:
• gid – original road ids in ascending order
• nodes – geometry column containing nodes as POINT geometry
Display
Shows the
path
Layer displays
the map network
Map Layer
Map
Nodes
Layer
Evader
path layer
Pursuer
path layer
Layers
(vector)
Evader
end point
layer
Shows the
current
positions
Pursuer
start point
layer
Evader
start point
layer
Pursuers
end point
layer
Layer displays
nodes in the map
Shows the
starting
positions
Results: capture time vs. number of
pursuers


Map used: Hyderabad Road Network
Difficulty of taking into account all the factors
responsible in chase is avoided by measuring
the simulation time over 10 and 20 simulation
runs and averaging the results
Result: measured over 10 simulation
runs
Results: measured over 20 simulation
runs
Future Work




Developing heuristic based strategies for both pursuer
and evader
Incorporating the road constraints
Automating all the tasks in required in the preprocessing
for visualization
Use of Raster layers instead of vector layers for displaying
map will speed up the process
References
 [1] Theory and Applications of Graphs, chapter Pursuit-evasion in a
graph. Springer Berlin / Heidelberg, 1978.
 [2] A. AlDahak and A. Elnagar. A practical pursuit-evasion algorithm:
Detection and tracking. In Robotics and Automation, 2007 IEEE
International Conference on, pages 343 - 348, April 2007.
 [3] W. Herbert and F. Mili. Route guidance: State of the art vs. state of
the practice. In Intelligent Vehicles Symposium, 2008 IEEE, pages
11671174, June 2008.
 [4] L. J. Guibas, J.-C. Latombe, S. M. LaValle, D. Lin, and R. Motwani.
A visibility-based pursuit-evasion problem. In Intl. J. of
Computational Geometry Applications, volume 9, pages 471 - 493,
1999.