Wireless Communication and Sensornets Laboratory

Department of Electrical and Computer Engineering
University of California, Santa Barbara

Home People Research Publications Collaborators Photos Resources Contact

Current research emphases

Research projects: quick summaries




  • Networked Estimation with Multimedia Sensors

    The availability of cheap and powerful multimedia sensors like cameras and microphones means that (a) we can deploy a lot of sensors and (b) record "high bandwidth" events, leading to lots of spatial and temporal correlation in the data that we gather. The key question that we would like to answer is: can we leverage the correlations in space and time, to build a sensor network that will "learn" to detect "interesting" events (bird calls/gunshots) with minimal prior information? Post-detection, we would also like to pool the sensor data to reconstruct "event signatures" with high fidelity.

    The main problem is to provide a definition of an "interesting" event. The approach we have taken so far is to decide purely based on the correlation between readings at neighboring sensors: an interesting signal is present if neighboring sensors show correlated readings, whereas uncorrelated sensor readings correspond to noise (picture). We propose detection and estimation algorithms for a setting where the sensors receive scaled and delayed versions of a common signal. Our key contributions are as follows:

    1. In the case when the received signals are time synchronized, we propose a GLRT detection rule and demonstrate that we can gain exponentially (in terms of probability of miss) by increasing the number of sensors. We also outline the ML estimation algorithm to reconstruct event "signatures".
    2. We then propose an algorithm to estimate the relative delays between readings at sensors that works well even at low SNR.
    3. Finally, we propose a distributed version of the detection algorithm that cuts down on the communication costs.
    You can check out the performance of our algorithm by comparing the noisy signal obtained at a single sensor and the reconstructed version.





    Students

    Sriram Venkateswaran

    Faculty

    Upamanyu Madhow

  • Joint ADC Compensation & Demodulation: Towards all Digital Baseband Architectures for Gigabit Communications

    All digital Gigabit baseband architectures need ADCs of sufficient rate and precision. The precision requirements are increased in case multiple antennas are used to obtain spatial multiplexing, higher constellations are used for spectral efficiency or digital equalization is done to combat dispersion. Since these scenarios arise in the communication protocol design for the 60GHz unlicensed band, ADCs with high precision (8-10 bits) sampling at 5GSa/s are needed. Time-interleaving several low rate ADCs is a good option to meet these requirements. This architecture can also result in power reduction due to the use of power-efficient low rate ADCs (SAR). But this comes with the issue of mismatch between the interleaved ADCs. To the first order, the mismatch can be classified as gain, timing and voltage offset mismatch. Present day mismatch compensation algorithms have high computational complexity. We have tried to address this issue by compensating for the mismatch jointly with the channel equalization. Mismatch is estimated using the already available channel training sequences and then compensated. Results for OFDM system employing Time-interleaved ADC indicate significant computational gains which we attribute to the special structure of mismatch interference in the frequency domain. Current research focuses on exploring such gains for MIMO systems in the 60GHz band and optical transceivers with dispersion. The challenges are the integration of mismatch compensation with the space-time equalization in MIMO systems with the compensation of chromatic dispersion in the optical transceivers and how we can scale the compensation algorithms with the number of interleaved ADCs.





    Students

    Sandeep Ponnuru

    Faculty

    Upamanyu Madhow, Mark Rodwell

    Collaborators

    Munkyo Seo

  • Signal Processing with Low Precision ADC : Towards Low Cost Gigabit Wireless Communication

    As communication systems scale up in speed and bandwidth, the power consumption and cost of high-precision (e.g., 6-12 bits) analog-to-digital conversion (ADC) becomes the limiting factor in modern receiver architectures based on digital signal processing. In this work, we are considering the effects of lowering the precision of the ADC on the performance of the communication link. The topics under investigation range from the design of DSP algorithms that explicitly take ADC imperfections into account for operations such as synchronization, channel estimation and equalization, to the derivation of information-theoretic limits that provide performance benchmarks for communication with imperfect ADC. Apart from transceiver design for high speed wireless communication, the general framework of signal processing with sloppy ADC could also have other potential applications, such as providing methods for monitoring wide swaths of spectrum for cognitive radio.

    Students

    Jaspreet Singh

    Faculty

    Upamanyu Madhow

    Collaborators

    Onkar Dabeer

    More Information



  • Millimeter-Wave MIMO: Wireless Links at Optical Speeds

    The millimeter-wave (mm-wave) MIMO architecture is designed to achieve optical data rates (40 Gbps and higher) over point-to-point wireless links. By allowing high-capacity wireless links to be set up quickly and cost-effectively, mm-wave MIMO has application in communication infrastructure recovery in the case of emergency or natural disaster. Additionally, the architecture allows inexpensive deployment of wireless "bridge" links between optical networks in environments where installation of fiber is difficult or costly (i.e. city centers, mountains, rivers, etc.).

    Mm-wave MIMO employs mm-wave spectrum in the E-band (71-95 GHz) to achieve spatial multiplexing gains in line-of-sight environments. The primary signal processing tasks are divided into an efficient two-level hierarchy. Transmit and receive beamforming, constituting the first level, provides high directivity and allows a range on the order of kilometers in adverse weather conditions. The second level consists of spatial equalization which allows parallel transmission of multiple multi-Gigabit-per-second data streams across the link. Operation at these high data rates presents a number of significant design challenges, demanding the use of hybrid analog/digital signal processing algorithms which are co-designed with the hardware. Hardware-oriented mm-wave MIMO research is currently being pursued by the research groups of Professors Mark Rodwell and Patrick Yue of UC Santa Barbara.



    Students

    Eric Torkildson, Colin Sheldon

    Faculty

    Upamanyu Madhow, Mark Rodwell, Patrick Yue

    Collaborators

    Raghuraman Mudumbai, Munkyo Seo

    More Information



  • Medium Access Control for 60 GHz Outdoor Mesh Networks

    This project investigates an architecture for multi-Gigabit outdoor mesh networks operating in the unlicensed 60 GHz millimeter (mm) wave band. In order to overcome the higher path loss at mm wave carrier frequencies, the use of narrow antenna beams is essential for attaining the required link ranges (of the order of 100 meters) with the transmit powers possible with low-cost silicon implementations. However, highly directional links make coordination between neighboring nodes for medium access control (MAC) difficult, since standard protocols such as carrier sense multiple access that rely on neighboring nodes hearing each other, become inapplicable. We explore the extent to which we can reduce, or even dispense with, coordination, exploiting the reduction in interference due to the narrow beamwidths and the oxygen absorption characteristic of the 60 GHz band.

    Students

    Sumit Singh

    Faculty

    Upamanyu Madhow, Elizabeth Belding, Mark Rodwell

    Collaborators

    Raghuraman Mudumbai, Munkyo Seo

  • Millimeter Wave WPAN: Cross-Layer Modeling and Multihop Architecture

    The 60 GHz band has been allocated worldwide for short range wireless communications because high atmospheric path loss due to oxygen absorption renders it unsuitable for long distance communications. This abundant unlicensed spectrum in the 60 GHz millimeter (mm) wave band offers the potential for multiGigabit indoor wireless personal area networking (WPAN). With the relentless scaling of the silicon semiconductor processes, low-cost transceiver realizations are within reach. However, mm wave communication links are more fragile than those at lower frequencies (e.g., 2.4 or 5 GHz) because of larger propagation losses and reduced diffraction around obstacles. On the other hand, directional antennas that provide directivity gains and reduction in delay spread are far easier to implement at mm-scale wavelengths.

    We are working on cross-layer modeling methodology and a novel multihop medium access control (MAC) architecture for efficient utilization of the 60 GHz spectrum, taking into account the preceding physical characteristics. Our in-room WPAN architecture constrains every link to be directional for improved power efficiency (due to directivity gains) and simplicity of implementation (due to reduced delay spread). We show that multihop relay to route around blocked links can assure consistent network connectivity even in the face of challenging scenarios with multiple stationary or moving obstacles. We have developed an elementary diffraction-based model to determine network link connectivity, and have defined a multihop MAC protocol that accounts for directional transmission/reception and includes procedures for topology discovery and recovery from link blockages.

    Students

    Sumit Singh, Federico Ziliotto

    Faculty

    Upamanyu Madhow, Elizabeth Belding, Mark Rodwell

    More Information



  • A Resource Biasing Framework for Shaping Throughput Profiles in Multihop Wireless Networks

    Throughput performance of multihop wireless networks is governed by how the network transport capacity is partitioned among different competing network flows. Max-min fair allocation leads to poor throughput performance for all flows because connections traversing a large number of hops consume a disproportionate share of resources. While proportional fair allocation provides a significant improvement, there is a much richer space of resource allocation strategies for introducing a controlled bias against resource-intensive long connections in order to significantly improve the performance of shorter connections. Our simple analytical model offers insight into the impact of a particular resource allocation strategy on network performance while also capturing the effect of finite network size and spatial traffic patterns. Our simulation results demonstrate that ''mixed'' bias strategies that blend fair allocations with a strong bias against long connections can provide significantly better performance to shorter connections than max-min fair or proportional fair resource allocations, with minimal impact on the performance of long connections.

    Students

    Sumit Singh

    Faculty

    Upamanyu Madhow, Elizabeth Belding

  • Sticky CSMA/CA: Implicit Synchronization and Real-time QoS for Mesh Networks

    We consider the problem of efficiently supporting a mix of real-time and delay-insensitive traffic over wireless mesh networks, assuming a narrowband physical layer with CSMA/CA capabilities. Classical CSMA/CA, as in IEEE 802.11, does not provide the delay guarantees required for real-time traffic. We introduce Sticky CSMA/CA, a MAC scheme that provides TDM-like performance without requiring explicit synchronization. We exploit the natural or artificially imposed periodicity of real-time flows to obtain implicit synchronization for efficient channel use and QoS guarantees. Nodes monitor the medium using the standard CSMA/CA and record the recent history of medium activity. A new real-time flow uses this information to grab the medium at the first available opportunity, and then sticks to a periodic schedule, providing delay and bandwidth guarantees. Delay-insensitive traffic fills the gaps left by the real-time flows. Large gains over standard CSMA/CA are demonstrated for a voice/data network.

    Students

    Sumit Singh, Prashanth Acharya

    Faculty

    Upamanyu Madhow, Elizabeth Belding

    More Information



  • Target Tracking with Binary Proximity Sensors

  • Fundamental Localization Limits of Binary Sensors

    We first explored the fundamental limits of localization using the simple model of binary proximity sensors (or "detection sensors"). Each sensor can only detect the presence or absence of a target within its sensing range, and nothing more. These sensors can be considered as abstractions for the communication and sensing limits of real sensors. We showed that increasing the sensing range makes each sensor less accurate, but by collaborating with other sensors the overall localization error actually decreases. We also showed that the localization error is subject to a fundamental lower bound, and this lower bound also implies fundamental limits on the tracking capability of the network. In particular, we showed that it is only possible to track targets whose trajectories are "smooth", and conversely the output of any tracking algorithm can be "smoothened" without losing information.

    More recently we followed up on this previous result, and showed that the localization limits derived previously for the special case of an ideal binary sensor can be extended to arbitrary types of sensors. Our analysis of this problem is based on a simple hypothesis testing problem related to the localization problem. This analysis used a novel mathematical technique of averaging over random spatial placement that also proved to be useful in providing a simple characterization of interference in mm-wave networks.

    Students

    Raghuraman Mudumbai, Nisheeth Shrivastava

  • Tracking Multiple Targets Using Binary Sensors

    In this part of the project, we explored the use of binary proximity sensors for tracking (an unknown number of) multiple targets. We first addressed the problem of counting the number of targets based on a snapshot of the sensor readings. We obtained necessary and sufficient criterion for an accurate target count in a 1-d setting, and provided a greedy algorithm to determine the minimum number of targets that is consistent with the sensor readings. To track targets across snapshots, we proposed a particle filtering algorithm based on the Bayesian probabilistic framework. Under the assumption of temporal continuity of the target trajectories, our algorithm provided excellent performance in tracking the targets.

    Students

    Jaspreet Singh, Rajesh Kumar

    Faculty

    Upamanyu Madhow, Subhash Suri

    More Information



  • Coopertive Beamforming for Wireless Networks

    Cooperative communication techniques are based on the idea of trying to achieve MIMO-like gains in data rate by using single-antenna radios like distributed elements of a virtual multi-element antenna array. However it turns out that certain theoretical abstractions from the MIMO literature cannot be directly extended to cooperative communication using virtual MIMO arrays. One such abstraction is the focus of this research i.e. the problem of synchronization.

    Traditional information theoretical analyses of wireless systems have implicitly assumed timing and carrier synchronization, by modeling the transmitted and received waveforms as complex baseband signals. However, for cooperative communication in a wireless network, this assumption completely breaks down; each transmitter and receiver uses its own local oscillator to obtain carrier signals, and offsets between these signals lead to unpredictable time-varying effects in the baseband signals.

    Our key contribution is a simple, feedback-based algorithm that completely avoids the need for explicit phase offset or channel estimation, by using a feedback control approach. In this algorithm, the cooperating transmitters make independent, random phase adjustments in each timeslot, and the receiver broadcasts a single bit of feedback whether or not the SNR increased compared to previous timeslots. This allows the transmitters to retain the "favorable" phase adjustments and discard the "unfavorable" ones and eventually achieves full beamforming gains. Also it can be shown that a substantial portion of the full beamforming SNR gains are obtained very quickly, and the algorithm is robust to noise and time-varying channels. Therefore this algorithm is well-suited for practical networks, as demonstrated in recent hardware prototypes at UC Berkeley and UCSB.






    Students

    Raghuraman Mudumbai, Ben Wild

    Faculty

    Upamanyu Madhow, Joao Hespanha

    Collaborators

    Kannan Ramchandran (UC Berkeley), Gwen Barriac (Qualcomm)

    More Information



Research Sponsors