Vehicles and pervasive computing

As you might now I attended PerCom a couple of months ago.

During the panel discussion, there was an interesting conclusion that vehicles are the most pervasive devices available today.
If you think about it vehicles contain a large number of sensors:

  • Speed
  • Acceleration
  • Yaw, G sensors (for ESP)
  • Temperature (environment, engine, tires)
  • Light sensors (automatic lights etc)
  • Fuel consumption, oxygen level, CO2 levels
  • GPS, Navigation system, maps etc.
  • Noise (to automatically regulate car radio volume)

Furthermore, modern vehicles have a number of ways to communicate (FM Radio (RDS,TMC), Bluetooth, GSM, soon 802.11n (WAVE) ). In my opinion, all these features already constitute vehicles as mobile sensor platforms. By exploiting already available sensors we can design numerous applications:

  • Road traffic monitoring (using acceleration and speed sensors)
  • Distributed pollution and temperature monitoring
  • Parking information
  • Formation of platoons of vehicles (e.g. maximize road capacity)
  • Dissemination of warning information
  • Accident avoidance (e.g. break when approaching a red light fast)
  • Visual enhancement (e.g. provide information on the windshield about traffic ahead, red light warning).
  • Landmarks/advertisement
  • Communication between vehicles (voice/file sharing etc)

All these application can be implemented either centrally (e.g. using GSM) or in an Ad-Hoc manner. Although the first approach is more reliable and fast it has some disadvantages:

  • Centralized data may be outdated and the response time may not meet the real-time requirements. Especially if we want local information (like “is the traffic light ahead red?”, “is there a parking spot within 100m?”)
  • Current centralized communication solutions (GSM, WiMAX) may not be able to handle the burden of real-time monitoring of hundreds of thousands of vehicles. These services are allready congested with million of mobile phone users.
  • Infrastructure could be quite expensive, especially if the area to be covered is large. Furthermore, infrastructure may not be available, especially in remote and isolated areas (you haven’t been on mountains in Greece ;) ).
  • Ad-hoc service is free and it can provide more concentrated local information (e.g. advertisements etc)

However, there are a lot of research issues in order to implement all these Ad-Hoc applications:

  • We need robust routing protocols that work both in dense urban areas and sparser areas (e.g. DTN).
  • We need dissemination protocols that take into account the interest of the vehicles/drivers to avoid flooding the drivers with unnecessary information and cause congestion to the network
  • We need MAC protocols that are able to deliver information to high-speed moving vehicles.
  • We need extremely robust trust and security mechanisms because there are human lives at risk!!

My last year’s research was manly focused on two areas:
How to exploit the navigation system of the car to route and disseminate messages, and how to use the Publish/Subscribe communication paradigm to achieve that.

Navigation systems are becoming more and more popular. A part from navigation suggestions, navigation systems provide valuable information that is currently not generally used. First of all, the GPS unit provides the vehicle’s geographical location. Furthermore, the NS map provides various geographical information: street names and numbers, location of points of interest (like fuel stations, train stations, etc), kilometer ranges, etc. This information may be extended to include the location of known infostations so that the vehicle can geographically route information to them.

At the same time, the most important information that the NS provides is the suggested route of the vehicle: This information 1)makes the mobility patterns of the vehicles more predictable. Additionally, 2)it can be exploited to extract interests, which can be used to design efficient routing and dissemination protocols and to filter information that is only relevant to the driver without his/her intervention (for example an accident warning affects only vehicles that will drive near the accident).

Therefore, in the last months I examined how we exploit the navigation systems (suggested routes to predict mobility patterns and to extract subscriptions) in order to design a vehicular routing algorithm and a Publish/Subscribe system.

You can find the results here http://www.cs.ucl.ac.uk/staff/I.Leontiadis/pubs.html
There is also some additional work on Pub/sub issues but we have to wait the reviewers first :)

Ilias Leontiadis

Comments are closed.