Home » Research » Contributions – Molecular Communication

Contributions – Molecular Communication


This page summarizes the problems that I have studied on molecular communication (MC), including those presented in my PhD thesis. The content on this page is organized by topic and not by paper to more easily clarify the contributions to different problems. To just see the list of publications, please refer to the Publications page. The publications here are referenced with the same numbering as on the Publications page, and the prefix corresponds to the type of paper (“J” for journal, “C” for conference, and “U” for unpublished).

The content on this page is current as of February 2017.


My research has focused on the study of communication via molecular diffusion, whereby molecules propagate via collisions with other molecules. Diffusion is ubiquitous in nature for communication, due to its simplicity and its speed over short distances. Systems that use molecular communication via diffusion have several characteristics that distinguish them from conventional wireless or wireless communication systems:

  1. Molecules are discrete objects that need to be collected or generated and then physically transported. These tasks take non-negligible time.
  2. Channel impulse responses, i.e., how a receiver observes an impulse of molecules released by a transmitter, are determined by solving differential equations with the environment’s boundary conditions. Closed-form solutions may not be available in many cases.
  3. Sources of uncertainty include the number of molecules present, the trajectory of each individual molecule, and other phenomena that can transport or transform molecules (e.g., chemical reactions or bulk fluid flow).
  4. Devices that communicate via diffusion, e.g., living cells, tend to be very small and have limited computational resources.

These characteristics demonstrate that the effective application of communications theory to molecular communication systems is unintuitive, despite the fact that the environments of interest have been studied for decades in the life and physical sciences. Interdisciplinary expertise is crucial for success in this domain. I developed my capacity to innovate by complementing my background in communications engineering with courses in cell biology and physical chemistry. I used these courses to guide me through independent learning of biochemical systems and existing simulation methods. From this foundation, I sought to improve the understanding of molecular communication channels, measure practical communications performance, and improve the availability of suitable simulation tools.

Understanding the Molecular Communication Channel

As an emerging domain, there are significant opportunities to guide the direction of molecular communication research by establishing fundamental knowledge. A communications channel that is based on molecular diffusion is clearly distinct from one that is based on conventional telecommunications technology. I have worked to enhance the understanding of diffusion-based molecular communication channels. This has included a rigorous study of the assumptions that are made to obtain tractable results, demonstrations of how to physically change the channel to improve the capability to communicate, and exploration of how a device might be able to estimate the system parameters. I discuss all of these contributions in the following:

  1. Verifying Analytical Assumptions. Meaningful transceiver system design and performance analysis relies on the availability of an expression for the channel impulse response. However, closed-form expressions for the impulse response of a diffusive channel generally require simplifying assumptions and specific system geometries. Limited efforts have been made to assess the validity of some of the most common simplifying assumptions. I have studied these assumptions in detail in order to demonstrate the applicability of any corresponding analysis. They include:
    • For tractability, it is often assumed that transmitters and receivers can be represented as points in 3D space. In [C12] and [C3], we derived the channel impulse response due to a spherical transmitter and at a spherical receiver, respectively, and measured the deviation from the simplified point transceiver model. The point assumptions are only valid if the distance between the transmitter and receiver are sufficiently small relative to their size.
      Uniform concentration assumption
      Deviation from the uniform concentration assumption as a function of time when a source releases an impulse of molecules into unbounded 3D space and the observer is a sphere. Each curve corresponds to a different receiver radius when the distance from the center of the sphere to the molecule source is 1. Source: [C3].
    • The statistics of a channel impulse response are critical for evaluating communications performance. It is generally assumed that observations at a receiver can be described by a random variable with a deterministic and time-varying mean, and furthermore that each observation is an independent realization of that random variable. In [J4], we derived the mutual information between observations of the number of molecules inside a transparent receiver. Our results demonstrated that observations in rapid succession are not independent, and this leads to overestimating the performance of detectors that rely on rapid sampling.
    • It is often assumed that the intended transmitter is the only source of molecules. However, in a biochemical environment, different types of molecules are often re-used for different purposes, so molecules at a receiver may arrive from different sources. We derived the number of molecules observed from continuously-emitting noise sources in [J5]. We showed in [C4] that such sources can have a potentially devastating impact on communication.
    • The observation of molecules by a receiver is often modeled as an ideal process. In practice, molecule detection could involve one or more chemical reactions that occur inside or on the surface of the receiver. In [C15], we derived transform functions between the two common ideal receiver models, thereby unifying their analysis. In [C11] and [J8], we derived the channel impulse response for a receiver that detects molecules via a reversible surface reaction, which is a more general and realistic model.
  2. Modifying the Communications Channel. One of the biggest challenges of using diffusion for communication is that the molecule propagation time increases with the square of the distance. Furthermore, the channel impulse response has a heavy tail. So, molecules can take a long time to arrive at their destination and they can continue to arrive over a long period, resulting in significant intersymbol interference (ISI). However, I demonstrated how the inclusion of other phenomena can mitigate the potential impact of ISI, as follows:
    • We drew inspiration from the confined environment in the neuromuscular junction, where information molecules diffuse from a neuron to a muscle fiber. There, enzyme molecules bind to and then degrade the information molecules so that they do not interfere with future signaling. We identified this as an example of ISI mitigation, and in [C2] and [J3] we derived a bound for the channel impulse response in a diffusive environment with enzymes present. Adding enzymes to the propagation environment was shown to significantly reduce ISI and thus improve the performance of simple detectors.
      Enzyme vs no enzyme
      Comparing the expected diffusion wave with and without molecule degradation via enzymes. The enzymes result in a weaker signal but with much less intersymbol interference. Source: [J3].
    • A fluid flow, such as in the bloodstream, imposes a net displacement on diffusing molecules. This can help to both carry molecules to their intended destination and then carry them away. In [J4][J5] and [C5], we derived the channel impulse response when there is a steady uniform flow in an arbitrary direction. We demonstrated that a flow can be too fast to be helpful for detection, and that even a slow flow in the direction opposite that from the transmitter to the receiver can actually improve communication.
  3. Estimating Channel Parameters. Analytical expressions for the channel impulse response are functions of the underlying channel parameters, such as the distance and the time since the molecules were released. Certain applications, such as health care diagnostics and advanced receiver designs, may need to monitor changes in these parameters. Therefore, observations of molecules can be used to make estimates of parameter values. In [C7] and [J6], we formulated the general channel parameter estimation problem for diffusive molecular communication and we derived a lower bound on the variance of unbiased estimation of any of the parameters. We demonstrated when maximum likelihood estimation can achieve the lower bound and proposed low-complexity methods for practical estimation.
    Accuracy in distance estimation
    Normalized error when estimating the distance from a transmitter by sampling the wave due to a single impulse. Error is written as a function of the number of samples. The error increases as we remove knowledge of the system parameters (shown is the loss of the release time and then the components of the steady flow vector). Source: [J6].

Communications Performance

Once we have established how to describe the molecular communication channel, we can consider how devices transmit information over the channel. For practicality, I have focused on a simple transmission scheme where a transmitter releases an impulse of molecules to transmit a bit 1 and no molecules to transmit a bit 0. This strategy, known as ON/OFF keying modulation, can be used to transmit arbitrary quantities of information that can be represented with binary digits. Given the transmission scheme, I have sought to design and analyze simple detection methods and compare their performance with optimized strategies. I began with a single communication link, with one transmitter and one receiver, and then expanded my study to networks with multiple transmitters and/or receivers, as follows:

  1. Single Transmitter-Receiver Design. In the single link case, there is only one destination and interference from any other links can be ignored. I made the following contributions to receiver design and analysis, which form the foundation for studying more complex networks:
    • In [J4], we proposed the maximum likelihood detector for sequential detection of the transmitter bits. The performance of this detector, which can be very good when the receiver takes multiple samples in every symbol interval, serves as a lower bound on the performance of more practical receiver designs. We also proposed weighted sum detection as a generalization of the existing single sample detector, and derived the corresponding bit error probability. The weighted sum detector acts as a filter by assigning a weight to every sample in a symbol interval and then comparing the sum of the weighted samples to a decision threshold. We showed that using matched filter weights in the absence of ISI had performance comparable to that of the maximum likelihood detector, and that using equal weights (i.e., an energy detector) also had good communications performance.
    • In the presence of ISI, the performance of weighted sum detectors is limited. One strategy to improve this is to use longer bit intervals, but this reduces the data transmission rate. Alternatively, we proposed variations of decision feedback equalization in [C4], [C16], and [J7], where the receiver uses its previous decisions to predict the level of ISI and adapts the weighted sum detector accordingly. We showed that feedback could improve the bit error probability considerably, albeit with a corresponding increase in computational complexity.
  2. Molecular Communication Networks. Examples of molecular communication in nature often involve many transceivers in the same area, such as cells in a tissue or a community of bacteria. Similarly, the envisioned applications of synthetic molecular communication networks would likely involve deployments of large numbers of devices. Thus, it is important to study how the presence of multiple transceivers might help or hinder the ability to communicate. I have studied several such systems, as follows:
    • Relaying, which is a common topic in communication networks, also plays an important role in biological systems to help signals propagate further and pass through physical barriers. In [C6], [C10], and [J7], we derived the bit error performance of a system that uses sequential “hops” between intermediate transceivers to transmit a data sequence. We modeled both a decode-and-forward (DF) scheme, where each intermediate node relayed a local bit decision, and an amplify-and-forward (AF) scheme, where each node relayed a multiple of its local observations. We identified the practical issues of interference when the same type of molecule is used for multiple hops, and proposed adaptive methods to mitigate the effects. Overall, we showed that the deployment of relays can significantly improve the end-to-end communications performance.
    • DF and AF relaying can be considered serial relaying, since there is only one “path” from the transmitter to the intended receiver. We also proposed parallel relaying, where the local decisions of multiple receivers are combined by a fusion center to make global decisions about a transmitter’s bit sequence. In [C13] and [U2], we derived the bit error performance of such cooperative detection systems using various “voting” rules, where a bit is decided based on the number of local receiver decisions. Furthermore, we optimized the decision thresholds at both the local receivers and the fusion center. We showed that cooperative detection can provide a significant improvement over the performance of a single link.
    • The presence of multiple receivers does not always guarantee an improvement in performance. For example, in [J9], we considered two independent receivers that absorb molecules, and we measured the degradation in detector performance at one receiver as a function of the location of the second receiver. We showed that having another absorber in the vicinity of the intended receiver or the molecule source can severely impact the communication performance.
    • Systems may have a large number of devices, and furthermore, due to their mobility, their locations may be random. In [C14] and [U1] (not publicly available), we applied stochastic geometry to derive the channel impulse response at a receiver due to simultaneous transmission by a field of randomly-placed transmitters. We also demonstrated the challenges in trying to demodulate a signal from such a network of transmitters, since there is a very large variance in molecule arrival times.

Simulation Methods

Simulation platforms can provide an efficient and inexpensive method to test and verify molecular communication systems. However, the use of simulations is not widespread in this field. Existing generic reaction-diffusion simulators from the biophysics community have limited applicability to molecular communication systems, since they are not designed to accommodate the modulation of data sequences and they are not designed to generate channel statistics over a large number of independent realizations. Existing simulators from the molecular communications community have limited flexibility for a regular user to study new and different environments.

I have sought to bridge this gap by developing a simulation platform that is a generic reaction-diffusion simulator and is also designed for communications analysis. In [C8][C9], we presented work on simulating diffusion and chemical reactions within user-defined environments. From this foundation, I developed the AcCoRD simulator (Actor-based Communication via Reaction-Diffusion), which is publicly available as an open source project on Github. In [J12], we presented AcCoRD, discussed its algorithms, and verified its use as an accurate reaction-diffusion simulator for the analysis of molecular communication systems. AcCoRD is a “sandbox”-style simulator with flexible options for a user to define a custom environment with molecule sources, observers, and chemical reactions. AcCoRD will improve the accessibility of the molecular communication field, provide a platform to verify new analysis, and enable the exploration of channels that have not or cannot be precisely examined analytically.