This page has moved to http://mmi.tudelft.nl/~patrick/Projects/IDA/

red line

The Intelligent Driving Agent (IDA) Project

THIS PAGE DESCRIBES A FINISHED PROJECT AND IS NOT UPDATED ANYMORE! red line
Back to KBS

Introduction

The Intelligent Driving Agents (IDA) project is the project that I have been working on for my Master's thesis. The goal of the project is to design an intelligent agent that can control a vehicle and adjust to its environment. The idea is to use human driving behavior as a basis and design a number of simple behavior rules for the agent. Using adjustable parameters such as aggresiveness, desired speed, acceleration rate, gap acceptance etc., these agents can drive just like humans do.
By combining Rodney Brooks' subsumption architecture with a blackboard architecture the following behavior-based agent was constructed:

Intelligent driving agent design

Reasoning

The agent's reasoning process is as follows:
  1. The agent stores all incoming information from its sensors in its memory.
  2. The short-term planner calculates the expected position of moving objects.
  3. The behaviour rules use the sensordata and expected positions to generate several proposals for action.
  4. The arbiter chooses (a combination of) the best proposal(s).
  5. The best proposal is sent to the vehicle.
Agent reasoning process

Implementation

The agent model above was tested and implemented in a prototype simulation environment. The environment consists of multi-lane roads, intersections, traffic lights and multiple vehicles. On a Intel Pentium III, 450Mhz with 64 MB of RAM, we were able to run the simulator with up to 40 agents working in parallel in real-time. Adding more agents was possible but resulted in a slow- running simulation.
The prototype simulator Agent status window

Future work

At the moment, we are busy improving the simulator program. First, we are changing the current simulator into a distributed program to connect agents running on different computers over the network. This way we can run much more agents and create realisitic traffic jams. Second, we plan to expand the simulation environment with more objects, such as traffic signs, crosswalks, buildings, and parked vehicles. In addition, we will add new behavior rules to the existing agents to deal with these new objects.

For an extensive summary (5 pages) of the project read this paper

A more complete description can be found in my Master's thesis.









Related interesting sites:
  • The Smartest Project
  • GMD City Traffic

    Other traffic microsimulators:
  • Trafficware - SimTraffic
  • Carnegie Mellon - SHIVA
  • Los Alamos - TRANSIMS
  • University of Helsinki - HUTSIM
    		•

    Page Top