Radio environment maps and the cognitive VANET

Right now I'm working on mechanisms for expressing application-level quality-of-service (QoS) requirements in vehicular ad-hoc networks. First, let me state what I mean by QoS in a VANET setting and, more importantly, what sets it apart from how QoS is typically defined and used in more traditional networking scenarios.

The Wikipedia definition highlights what is typically meant by QoS:

"Quality of service is the ability to provide different priority to different applications, users, or data flows, or to guarantee a certain level of performance to a data flow"

In other words traditional QoS is all about moderating access to a shared communication channel in situations when the capacity of the link is not large enough to support all user requirements. Assumptions underlying this definition is that the channel itself is fairly static and that the dynamicity lies in the number and requirements of channel users. There's a few ways you can do this, admission control and data scaling comes to mind.

In the VANET case those assumptions no longer hold. For a mobile wireless network the main dynamicity lies in the channel (e.g. fading caused by obstacles) sometimes there will be no channel at all. In the VANET case the data traffic is highly predictable, most packets will be beaconing-messages to periodically inform other vehicles of your kinematic state. Thus, admission control and data scaling will not be enough to "guarantee" QoS (guaranteeing anything on a wireless link is an oxymoron).

What we can do however is adapt.

Recent measurements that we have made of microwave communication between vehicles indicate that there is a clear connection between the location of the transmitting and receiving nodes in relation to the environment (e.g. buildings and terrain). Since a VANET will consist of a large number of nodes travelling within fairly restricted areas (only on roads) we can use them as probes to measure communication quality as a function of their location.

This recorded data would provide us with the answer to one of the three questions that must be answered to implement a cognitive VANET (COG-VANET).

• "What do we have?" - What is the predicted communication quality based on node observations?
• "What do we need?" - We need a specification of application requirements expressed in a suitable way.
• "What should we do?" - There needs to be a strategy when requirements cannot be met, this will happen frequently and the services provided by the VANET must remain safe.

We've already outlined our work on the "What do we need?" question in a recent paper (I'll link it as soon as it's published) where we define a nifty way of expressing VANET application requirements. Currently I am working on the "What do we have?" question by gathering real world data, finding a good representation and a mechanism for comparing it to the requirements specification.

If you are working on or are interested in collaborating in this direction feel free to get in touch, I'm actively looking for research/industry internship opportunities as part of my PhD education.

Radio Environment Maps for VANETs?

The unreliable wireless link is perhaps one of the biggest obstacles to implementing cooperative safety functions using V2V communication. At 5.9GHz, the standardized frequency band for ITS applications in many countries, the effects of for example shadowing could be severe. Sure, more advanced radios using better coding techniques could help in a lot of scenarios, a while ago we wrote about such tech from Australian company Cohda Wireless. However good there will most likely still be corner cases where features of the terrain or buildings conspire to create radio "black-spots".

What I am interested in is to let the network learn by observing and recording communications quality as a function of location. This information would then be used to allow app developers to place requirements on communication quality. This idea is not completely new though, in the cognitive radio community adapting to the observed state of the wireless medium is one of the core tasks.

A quick literature search turns up quite a lot of publications on something called "Radio Environment Maps" (REM) by researchers at Virginia Tech. A REM in this context is basically a database of everything of interest to a cognitive radio, such as spectrum regulations, locations of primary users as well as past experience of nodes in the network. Although the authors mention several applications I believe they have the classical cognitive radio application in mind, spectrum scavenging. In the VANET case, the functionality of the existing (fixed spectrum) communications technologies as a function of location would be in focus.

The VANET REM I envision would have to be very detailed, i.e. so that communication quality between relevant pairs of locations could be looked up. My idea is to piggyback the data collection on top of the standard periodic beaconing protocols suggested for VANETs. An aggregation infrastructure could then be used to build the maps themselves.