Travel Report from University of Illinois

During September 2001 I visited University of Illinois at Urbana Champaign (usually referred to as UIUC, or just U of I), and the Coordinated Science Laboratory (CSL). My host for the visit was professor Kumar, who is a professor in control and communication. The visit was paritally funded by ARTES travel support.

CSL hosts professors in various fiels, such as communication, control, computer science, mechanics, etc., to support cooperation and inter-diciplinary research. This works to some extent (some groups have rather overlapping research).

Kumar's group works primarily with wireless network problems. Some students work on new wireless MAC protocols (to share the medium), others work on capacity theorems for wireless data. One of the more famous results from the group is that the capacity of wireless networks with N nodes scales as 1/sqrt(N) bit-meters/second/user, regardless if optimal or random positions are used. Thus, the more users, the less capacity per user. The idea of a cost of area around the receiver for correct reception is central.

Many practical experiments are performed by the group to back-up the theory. They have some 30 Linux laptops with Cisco 802.11 wireless LAN cards spread out in the building, to run capacity and routing experiments. The 802.11 standard is nice to experiment with due to its simplicity (all users in the same "channel", virtually replaces the wired ethernet).

Using the same wireless cards, a new lab has been set up to experiment with complex wireless systems. About 20 radio-controlled cars have been bought, and they intend to control these using WLAN with camera feedback. With this they can experiment with e.g traffic control. More important for this project, though, is the software architecture. How does one build a system which is robust and general, with lots of sensors, actuators, safety nets and objectives spread out and all communicating?

Over to real-time issues; there will of course be control problems using the wireless networks. Signals get delayed or lost. For this reason I gave an overview seminar on the topic of control using networks with focus on delays, which was well received. Also wireless networking involves a lot of scheduling, especially for routing.

Overall, the three week visit was very rewarding, finding new friends in the area of networked control.




Bo Lincoln
Last modified: 2001-11-15