With footholds in both public and private sectors and strong interdisciplinary backgrounds, LIDS graduates have a wide array of career paths from which to choose. Despite the multitude of available options and the diversity of his experiences and research interests, LIDS alumnus Todd Coleman is unsurprised at the direction his life has taken.

“I always knew I was going to like engineering,” he jokes, “because I used to play with Legos a lot.” All joking aside, Todd seems perfectly content with his chosen path. On a recent visit back to Boston, he sat down with LIDS-All to discuss the evolution of his research, life after LIDS, and the value of simplicity.

Originally from Dallas, TX, Todd attended college at the University of Michigan-Ann Arbor, did internships on the West Coast, and completed his M.S. and Ph.D. at LIDS, before recently accepting a teaching position at the University of Illinois at Urbana-Champaign. As a self-proclaimed expert in the “idiosyncrasies” of different regions of the United States, Todd loves the “friendliness of people” in the Midwest and looks forward to teaching graduate students at UIUC. The first course he will teach, Stochastic Processes, represents a return to his roots as well. “It was the first class I took at MIT,” he says, describing the course content as “probability on steroids.” Teaching will also provide Todd with the opportunity to recruit a graduate research assistant to support his current research into methods for achieving better communication from “multiple information sources” to “multiple receivers” over wireless networks, using what he calls “low-complexity solutions.”

In plainer English, Todd addresses the communication problems that can occur in different wireless scenarios, such as mobile ad-hoc networks or sensor networks, and tries to develop simple yet effective ways of resolving these problems. He offers the hypothetical example of “a group of temperature sensors that would like to communicate their temperatures to a group of monitoring stations.” Wireless sensors, tiny devices that can sense and record different environmental factors, are typically dispersed throughout a geographic region to monitor specific environmental phenomena, such as heat, light, or humidity. Because the devices are so small, they have limited energy. To conserve power and so that scientists can make use of the data, the sensors transmit their findings to centralized monitoring stations.

So far, this sounds fine, but Todd points out some possible problems in this type of scenario. “Because the sensors may be located in close geographical proximity,” he says, “the temperature readings will be correlated. Any good transmission scheme must exploit this correlation to reduce inefficiency.” In other words, since the temperatures recorded by one sensor node may be the same or very close to the temperatures recorded by its neighboring node, it is important to limit the total amount of information transmitted, in order to conserve energy on both the transmitting and receiving ends of the transaction. One way of doing this is to use what Todd calls a “distributed data compression approach,” which means that each sensor node will separately compress and encode the information it wants to send, but “knowing that the decoder [at the other end] wants to reconstruct both.” This means that redundant information is not sent thereby saving time, money, and energy.

Another potential problem in scenarios like sensor networks is interference, or simply, the possibility that information may not reach its destination intact. Todd explains, “interference between the wireless transmissions of each sensor node creates a problem whereby transmissions from different sensors… can create ‘collisions’ where the information is not successfully decoded.” If information from many different sensor nodes is attempting to reach the same monitoring station at once, some of the information may be lost or unintelligible at the receiving end of the transmission. Reducing such interference is crucial to the improved design of sensor networks.. Todd’s research represents an interesting approach to such “interference management,” which relies on the basic concept of cooperation. As he explains, “it turns out that, in some scenarios, a small amount of cooperation between the sensor nodes can sometimes significantly aid in addressing this problem.” For example, if the sensor nodes are able to schedule transmissions that don’t interfere with one another, “collisions” of this sort can be avoided.

With each of the engineering problems Todd researches, he seeks to develop “low-complexity solutions,” meaning solutions that can be simply and practically applied with “provably good results.” Simplicity is an attribute he admires in his colleagues’ work as well. He mentions LIDS professors Médard and Gallager as inspirations in his field and praises Professor Tsitsiklis for his “unique, elegant, well thought-out, simple explanations.” Overall, Todd feels lucky to have been a beneficiary of what he calls the “LIDS perspective.” He says “the great thing about LIDS is the importance [placed on] looking at deep interesting research problems and not taking shallow perspectives…. it’s something that always made me proud to be a LIDS student.” That emphasis on acquiring fundamental insights into problems has helped Todd’s research interests evolve during his time at LIDS. “I try to shy away from problems that are very traditional, that a lot of smart people have worked on and not been able to solve,” he says. “I like to go on tangents.”

These tangents have even led Todd to a completely different field: computational neuroscience. After graduation, Todd worked as a post-doc at Brain and Cognitive Sciences, just across the street from LIDS but seemingly worlds away from information theory. Yet important connections exist between the two fields, and Todd is interested in their intersection. “When I first took information theory and they talk[ed] about transmission of information,” Todd says, “the first thought that came to my mind was the nervous system and how your brain uses neurons to communicate with your arms to move in one way or another. It was always in the back of my mind, trying to understand how information theory might relate to different biological processes and the nervous system.” In fact, Todd explains that we are already beginning to recognize ways in which information theory can provide insights to neuroscience. He suggests, “a good interface between information theory and neuroscience could lie at the boundaries between neuroscience and engineering” and offers the example of a person with a prosthetic limb. If that person wants to use neurosignals to operate the limb, LIDS-style engineering enters the picture. As Todd points out, “you need to develop algorithms to take the neurosignals and control the device. There’s actually true engineering taking place, real control theory and communication …. That’s where the fun LIDS stuff comes in.”

Despite Todd’s enthusiasm for the lab, he admits to being nervous before arriving at MIT. He says, “I went to a state school that had a whole lot more grass, and a football team that people didn’t have to be subpoenaed to watch. There were quite a few differences. But I was pleasantly surprised.” With his diverse interests and love for tangents, Todd Coleman can expect more of such pleasant surprises. His current teaching position at the University of Illinois is only the beginning of his future in academia, and Todd predicts that years to come will find him “in the lab, working with grad students, working on interesting problems.” He foresees “a few gray hairs” as well, and laughs as he runs his hand over his shaved head. “Maybe not,” he amends.