Following is a list of suggested projects, and a brief description. You are free to come up with your own idea for a project. Discuss them with me before finalizing.

1. Functionality/code migration for increased network lifetime
2. Smart radios for multihop relaying
3. Content adaptation for mulithop wireless multimedia traffic
4. Distributed leader election in ad hoc network
5. Evaluation of joint-source channel coding for multihop wireless
6. Radio baseband processing on an FPGA
7. Communication using ultrasound
8. Energy-aware control of variable spreading gain in radios
9. Evaluation of the interplay of routing and topology management in sensor networks
10. Payment strategies for ad hoc networks

1. Functionality/code migration for increased network lifetime
   
   Often times the same function can be performed at different places in a networked system (e.g. on a handheld vs. the basestation, or at any of a bunch of overlapping sensors). One can use this to change the mapping of function-to-node and thereby manage system-wide energy resources to increase the lifetime of the overall network. The change of mapping may involve actual migration of code, but more likely messages to activate functions at different nodes at different times. As for example, consider an environment where people wear a sensor badge that has a microphone, and the badges talk to a backend-infrastructure. Most of the times the speech data is used to do word spotting as part of a speech driven interface. But at times one wants the speech of specific people to be captured into a multimedia database. Then a good strategy would be to by default do frontend signal processing for speech recogntion on the badge and send the resulting vectors over the air thereby saving lots of energy. Only when one needs to capture speech does one need to send speech over the air. In that case, one can activate a speech compression coder on the  badge, and do the decompression as well as front-end signal processing for recognition on the infrastructure. As another example, one can vary the quality of sensor data over time (resolution, rate) to adapt to available energy to meet lifetime requirements. The essence here is to exploit the communication-computation energy trade-off. The project would be to explore this idea, perhaps by way of an implementation on a paltform such as a PDA with a radio and sensors talking to a backend infrastructure.
    
2. Smart radios for multihop relaying
   
   In multihop networks a large fraction of the traffic carried by nodes is the relay traffic - i.e. packets that they need to forward as opposed to packets destined for an application on the node. However, in typical systems we have a "dumb" radio connected to a "smart" processor. So what happens is that all packets, even if they are to be only forwarded, get sent to the processor which is usually a power hungry CPU. The goal in this project would be to see how to put minimal smartnesss in the radio hardware so that it can (a) classify packets to find those that are to be forwarded, and (b) modify certain fields before queuing them for transmission (e.g. desitnation and source MAC addresses, time-to-live field etc.). A possibility is to explore radios with FPGA h/w where the routing daemon on the processor downloads rules for packet classification (kind of like what happens in firewalls) and packet modification to the FPGA, and then the processor is put to sleep. If the packet just needs to be forwarded, the radio will handle it without waking the processor. The basic idea was explored by one of my student as part of his M.S.  but at a very high level.
    
3. Content adaptation for mulithop wireless multimedia traffic
   
   Imagine a multihop network where pairs of users are involved in audio or video commmunication. DUe to wireless and mobility issues, the capacity of the network and the quality of paths between specific pairs will vary over time. The goal here would be to explore how the networking layer interacts with smart codec that offer you knobs to adapt the quality of the content being transmitted. E.g. if the available bandwidth goes down, perhaps we can gracefully reduce the resolution of several stream instead of killing some of the users. The best approach here would be real implementation.
    
4. Distributed leader election in ad hoc networks
   
   In ad hoc networks such as senor nets a key primitive that is often needed is distributed leader election ... e.g. a bunch of sensors detect something and now need to elect a leader to combine their  separate readings. Leader election is a special case of the problem of distributed consensus - i.e. how can a bunch of uncoordinated nodes "agree" on something. The goals here would be to explore the various (very few) algorithms that have been proposed for this in the context of wireless ad hoc networks, evaluate them by implementing them in a common simulation platform (e.g. parsec), and perhaps come up with a better algorithm.
    
5. Evaluation of joint-source channel coding for multihop wireless
   
   A hot topic in recent years among coding researchers has been the notion of joint source channel coding. Source coding refers to compression (remove redundancy) while channel coding refers to forward error correction (add redundancy). Usually these are done separately (source coding is an application level thing while channel coding is a link or physical layer thing). Joint source channel coding advocates doing the two together and shows that there are gains to be had from this. The problem is that these researchers know little about real-life networks and don't adequately address the problems that arise from mixing and end-to-end service (source coding) with a link-level service (channel coding). This is not an issue if all you have is a single wireless link (e.g. cellular). But, what happens in a multihop network. If you do "joint coding", which channel do you code to decode for? Last link? First link? What if they are good but it is the intermediate link that is bad? Well, one could decode/re-encode at every intermediate node, but then the latency will kill you. Perhaps one could seek support from the network to identify the worst link (the bottleneck link) and do the joint coding for that? The goal here would be to see how existing joint source channel  codes perform under multihop scenarios and explore the last idea (network helps you in identifying the bottleneck link). This would be simulation project ...perhaps matlab with parsec/glomosim/ns2.
    
6. Radio baseband processing on an FPGA
   
   Implement a simple radio  baseband (e.g. QAM) on an FPGA, and perhaps mate it with an off-shelf RF board. This would be in context of "reconfigurable radios" whose modulation can be changed on the fly depending on the other end. Explore protocols for such reconfiguration.
    
7. Communication using ultrasound
   
   Many projects are using ultrasound for purposes of localization. Can we use the same link to actually send some data also? Implement a simple communication link using an ultrasound transmitter and receiver (hardware is there ... so this is software only).
    
8. Energy-aware control of variable spreading gain in radios
   
   Direct-sequence spread spectrum radios are a type of radios where each information bit of length T is mapped to a sequence of N "chips" of length T/N, with the sequence corresponding to a code assigned to the receiver. For example, each 1 might be represented by (say) 0101 while each 0 by ~0101 = 1010 with N=4. N is called the spreading gain, and indicates the factor by which the signal spectrum is spreaded. The robustness of the link is related to N as at the receiver the interferers are suppressed by this factor N, in essence giving a SNR boost of N. Therefore the name spreading gain. Some radios permit the spreading gain to be varied over time. Of course the transmitter and the receiver need to coordinate. By varying the spreading gain, with a given bandwidth channel, once essentially gets a rate vs. robustness tradeoff: as the spreading gain increases, the link becomes more robust (higher SNR) but the data rate goes down (the max chip rate is a constant). Note that the BER changes as SNR does. The goal of this project is to investigate strategies for controlling spreading gain with the goal of energy awareness - i.e. minimize energy per good user level bit transported across. Variable spreading gain has been considered for robusteness by a group at V. Tech. in WLAN context, and more superficially by some of my and Prof. Rajeev Jain's students
   in 1997-98 in context of a radio. Note that the benefit of variable spreading gain comes when the channel is time varying. So essentially, the strategy would be to have a technique to change the spreading gain as the channel varies. If there are timing constraints, things become interesting! The project would involve simulation. In addition, I have VHDL files for a spread spectrum transceiver, that you can use to obtain data on how computational energy varies as you change the spreading gain.
    
9. Evaluation of the interplay of routing and topology management in sensor networks
   
   In the class I've described that for short range radios the way to save power is radio shutdown as the electronics power dominates. With this observation some of my students have developed a system whereby the radios in a sensor network are shutdown, and when an event happens, a "wakeup channel" is used to activate sleeping radios along the path. The STEM idea is described in
   C. Schurgers, V. Tsiatsis, and M.B. Srivastava, "STEM: Topology management for  energy efficient sensor networks," IEEE Aerospace Conference, March 2002. at http://nesl.ee.ucla.edu/pw/nesl/. (there are more recent papers that I can give). However, the work thus far has assumed an orthogonality with routing protocols... but clearly some routing protocols will work better than others. E.g. routing protocols requiring periodic wakeup or requiring flooding can negate the benefits of STEM. The goal of this project is to study the interaction of STEM and such topology management strategy with ad hoc routing protocols, and evaluate the total energy efficiency. In particular, some geographical version of Dynamic source Routing or a version of diffusion routing are possible candidates. The STEM system is currently implemented in PARSEC. So the project will involve mating it with one or more routing protocols, evaluating the impact, and changing STEM if needed to make it work better with the routing protocol.
    
10. Payment strategies for ad hoc networks
    
    This project seeks to address a key issue that needs to be resolved for the commercial viability of ad hoc networks. The issue is this: why would some one be willing to let their device be let used as a relay node to provide service to other traffic. Well, one possibility is economic inducement: nodes can earn credits for forwarding other traffic, and can use this credit later. Some nascent work has happened in this space (e.g. from Rutgers). Also, peer-to-peer networks such as gnutella, morpheus, kazaa etc. might be using some mechanism too. The goal of this project would be to see what techniques exist, and comparatively evaluate them using a simulator. Also, perhaps come up with some new strategy (or modification to existing one)

Back to EE206A Home Page
This page was last modified April 14, 2002
mbs@ee.ucla.edu