Spectral machine vision collects both the spectral and spatial dependence (x,y,λ) of incident light, containing potentially useful information such as chemical composition or micro/nanoscale structure.  However, analyzing the dense 3D hypercubes of information produced by hyperspectral and multispectral imaging causes a data bottleneck and demands tradeoffs in spatial/spectral information, frame rate, and power efficiency. Furthermore, real-time applications like precision agriculture, rescue operations, and battlefields have shifting, unpredictable environments that are challenging for spectroscopy. A spectral imaging detector that can analyze raw data and learn tasks in-situ, rather than sending data out for post-processing, would overcome challenges. No intelligent device that can automatically learn complex spectral recognition tasks has been realized.       UC Berkeley researchers have met this opportunity by developing a novel photodetector capable of learning to perform machine learning analysis and provide ultimate answers in the readout photocurrent. The photodetector automatically learns from example objects to identify new samples. Devices have been experimentally built in both visible and mid-infrared (MIR) bands to perform intelligent tasks from semiconductor wafer metrology to chemometrics. Further calculations indicate 1,000x lower power consumption and 100x higher speed than existing solutions when implemented for hyperspectral imaging analysis, defining a new intelligent photodetection paradigm with intriguing possibilities.

      Probabilistic Seismic Hazard Analysis (PSHA) has become a foundational method for determining seismic design levels and conducting regional seismic risk analyses for insurance risk analysis, governmental hazard mapping, critical infrastructure planning, and more. PSHA traditionally relies on two computationally intensive approaches: Riemann Sum and conventional Monte Carlo (MC) integration. The former requires fine slices across magnitude, distance, and ground motion, and the latter demands extensive synthetic earthquake catalogs. Both approaches become notably resource intensive for low-probability seismic hazards, where achieving a COV of 1% for a 10−4 annual hazard probability may require 108 MC samples.       UC Berkeley researchers have developed an Adaptive Importance Sampling (AIS) PSHA, a novel framework to approximate optimal importance sampling (IS) distributions and dramatically reduce the number of MC samples to estimate hazards. Efficiency and accuracy of the proposed framework have been validated against Pacific Earthquake Engineering Research Center (PEER) PSHA benchmarks covering various seismic sources, including areal, vertical, and dipping faults, as well as combined types. Seismic hazards are calculated up to 3.7×104 and 7.1×103 times faster than Riemann Sum and traditional MC methods, respectively. Coefficients of variation (COVs) are below 1%. Enhanced “smart” AIS PSHA variants are also available that outperform “smart” implementations of Riemann Sum by a factor of up to 130.

Optical atomic clocks have taken a giant leap in recent years, with several experiments reaching uncertainties at the 10−18 level. The development of synchronized clock networks and transportable clocks that operate in extreme and distant environments would allow clocks based on different atomic standards or placed in separate locations to be compared. Such networks would enable relativistic geodesy, tests of fundamental physics, dark matter searches, and more. However, the leading neutral-atom optical clocks operate on wavelengths of 698 nm (Sr) and 578 nm (Yb). Light at these wavelengths is strongly attenuated in optical fibers, posing a challenge to long-distance time transfer. Those wavelengths are also inconvenient for constructing the ultrastable lasers that are an essential component of optical clocks. To address this problem, UC Berkeley researchers have developed a new, laser-cooled neutral atom optical atomic clock that operates in the telecommunication wavelength band. The leveraged atomic transitions are narrow and exhibit much smaller black body radiation shifts than those in alkaline earth atoms, as well as small quadratic Zeeman shifts. Furthermore, the transition wavelengths are in the low-loss S, C, and L-bands of fiber-optic telecommunication standards, allowing the clocks to be integrated with robust laser technology and optical amplifiers. Additionally, the researchers have identified magic trapping wavelengths via extensive studies and have proposed approaches to overcome magnetic dipole-dipole interactions. Together, these features support the development of fiber-linked terrestrial clock networks over continental distances.

         Recent mass evacuation events, including the 2018 Camp Fire and 2023 Maui Fire, have demonstrated shortcomings in our communication abilities during natural disasters and emergencies. Individuals fleeing dangerous areas were unable to obtain fast or accurate information pertaining to open evacuation routes and faced traffic gridlocks, while nearby communities were unprepared for the emergent situation and influx of persons. Climate change is increasing the frequency, areas subject to, and risk-level associated with natural hazards, making effective communication channels that can operate when mobile network-based systems and electric distribution systems are compromised crucial.         To address this need UC Berkeley researchers have developed a mobile network-free communication system that can function during natural disasters and be adapted to most communication devices (mobile phones and laptops). The self-organized, mesh-based and low-power network is embedded into common infrastructure monitoring device nodes (e.g., pre-existing WSN, LoRa, and other LPWAN devices) for effective local communication. Local communication contains dedicated Emergency Messaging and “walkie-talkie” functions, while higher level connectivity through robust gateway architecture and data transmission units allows for real-time internet access, communication with nearby communities, and even global connectivity. The system can provide GPS-free position information using trilateration, which can help identify the location of nodes monitoring important environmental conditions or allowing users to navigate.

Silicon photonics is increasingly used in an array of communications and computing applications. In many applications, photonic chips must be coupled to optical fibers, which remains challenging due to the size mismatch between the on-chip photonics and the fiber itself. Existing approaches suffer from low alignment tolerance, sensitivity to fabrication variations, and complex processing, all of which hinder mass manufacture.To address these problems, researchers at UC Berkeley have developed a coupling mechanism between a silicon integrated photonic circuit and an optical fiber which uses a microlens to direct and collimate light into the fiber. Researchers have demonstrated that this device can achieve low coupling loss at large alignment tolerances, with an efficient and scalable manufacturing process analogous to existing manufacture of electronic integrated circuits. In particular, because the beam is directed above the silicon chip, this method obviates dry etching or polishing of the edge of the IC and allows the silicon photonics to be produced by dicing in much the same way as present electronic integrated circuits.

Measurement of electrical activity in nervous tissue has many applications in medicine, but the implantation of a large number of sensors is traditionally very risky and costly. Devices must be large due to their necessary complexity and power requirements, driving up the risk further and discouraging adoption. To address these problems, researchers at UC Berkeley have developed devices and methods to allow small, very simple and power-efficient sensors to transmit information by backscatter feedback. That is, a much more complex and powerful external interrogator sends an electromagnetic or ultrasound signal, which is modulated by the sensor nodes and reflected back to the interrogator. Machine learning algorithms are then able to map the reflected signals to nervous activity. The asymmetric nature of this process allows most of the complexity to be offloaded to the external interrogator, which is not subject to the same constraints as implanted devices. This allows for larger networks of nodes which can generate higher resolution data at lower risks and costs than existing devices.

Computer-based, internet-connected, audio/video meeting platforms have become pervasive worldwide, especially since the 2020 emergence of the COVID-19 pandemic lockdown. These meeting platforms include Cisco Webex, Google Meet, GoTo, Microsoft Teams, and Zoom. However, those popular platforms are optimized for meetings in which all the participants are attending the meeting online, individually. Accordingly, those platforms have shortcomings when used for hybrid meetings in which some participants are attending together in-person and others attending online. Also, the existing platforms are problematic for large meetings in big rooms (e.g. classrooms) in which most or all of the participants are in-person. To address those suboptimal meet platform situations, researchers at UC Berkeley conceived systems, methods, algorithms and other software for a meeting platform that's optimized for hybrid meetings and large in-person meetings. The Berkeley meeting platform offers a user experience that's familiar to users of the conventional meeting platforms. Also, the Berkeley platform doesn't require any specialized participant hardware or specialized physical room infrastructure (beyond standard internet connectivity).

The field of autonomous transportation is rapidly evolving to operate in diverse settings and conditions. However, as the number of autonomous vehicles on the road increases the complexity of the computations needed to safely operate all of the autonomous vehicles grows rapidly. across multiple vehicles, this creates a very large volume of computations that must be performed very quickly (e.g., in real or near-real time).   Thus, treating each autonomous vehicle as an independent entity may result in inefficient use of computing resources, as many redundant data collections and computations may be performed (e.g., two vehicles in close proximity may be performing computations related to the same detected object). To address this issue, researches at UC Berkeley proposed algorithms for the management and exchange of shared information across nearby and distant vehicles.According to the proposed arrangement, autonomous vehicles may share data collected by their respective sensor systems with other autonomous vehicles and adjust their operations accordingly in a manner that is more computationally efficient. This can not only increase safety but at the same time reduce computational load required by each individual vehicle.

The field of autonomous transportation is rapidly evolving to operate in diverse settings and conditions.  Critical to the performance of autonomous vehicles is the ability to detect other objects in the autonomous vehicle’s vicinity and adjust accordingly. To do so, many autonomous vehicles utilize a variety of sensors, including sonar. Although these sensor systems have been shown to improve the safety of autonomous vehicles by reducing collisions, the sensor systems tend to be computationally inefficient.  For instance, the sensor systems may generate large volumes of data that must be processed quickly (e.g., in real or near-real time).  The performance of excessive computations may delay the identification and deployment of necessary resources and actions and/or increase the cost of hardware on the vehicle making it less financially appealing to the consumer. Researches at UC Berkeley proposed algorithms for temporally and spectrally adaptive sonar systems for autonomous vehicles. These allow utilization of existing sonar system in an adaptive manner and in interface with existence hardware/software employed on autonomous vehicles. 

Electromagnetic noise from surfaces is one of the limiting factors for the performance of solid state and trapped ion quantum information processing architectures. This noise introduces gate errors and reduces the coherence time of the systems. Accordingly, there is great commercial interest in reducing the electromagnetic noise generated at the surface of these systems.Surface treatment using ion bombardment has shown to reduce electromagnetic surface noise by two orders of magnitude. In this procedure ions usually from noble gasses are accelerated towards the surface with energies of 300eV to 2keV. Until recently, commercial ion guns have been repurposed for surface cleaning. While these guns can supply the ion flux and energy required to prepare the surface with the desired quality, they are bulky and limit the laser access, making them incompatible with the requirements for ion trap quantum computing.To address this limitation, UC Berkeley researchers have developed an ion gun that enables in-situ surface treatment without sacrificing high optical access, enabling in situ use with a quantum information processor.

The real-time position and mobility of a user is key to providing personalized location-based services (LBSs) – such as navigation. With the pervasiveness of GPS-enabled mobile devices (MDs), LBSs in outdoor environments is common and effective. However, providing equivalent quality of LBSs using GPS in indoor environments can be problematic. The ubiquity of both WiFi in indoor environments and WiFi-enabled MDs, makes WiFi a promising alternative to GPS for indoor LBSs. The most promising approach to establishing a WiFi-based indoor positioning system requires the construction of a high quality radio map for an indoor environment. However, the conventional approach for making the radio map is labor intensive, time-consuming, and vulnerable to temporal and environmental dynamics. To address this situation, researchers at UC Berkeley developed an approach for automatic, fine-grained radio map construction and adaptation. The Berkeley technology works both (a) in free space – where people and robots can move freely (e.g. corridors and open office space); and (b) in constrained space – which is blocked or not readily accessible. In addition to its use with WiFi signals, this technology could also be used with other RF signals – for example, in densely populated and built-up urban areas where it can be suboptimal to only rely on GPS.

Capacitive-gap transduced micromechanical resonators routinely post Q several times higher than piezoelectric counterparts, making them the preferred platform for HF and low-VHF (e.g. 60-MHz) timing oscillators, as well as very narrowband (e.g. channel-select) low-loss filters. However, the small electromechanical coupling (as gauged by the resonator's motion-to-static capacitance ratio, Cx/Co) of these resonators at higher frequency prevents sub-mW GSM reference oscillators and complicates the realization of wider bandwidth filters. To address this situation, researchers at UC Berkeley developed a capacitive-gap transduced radial mode disk resonator with reduced mass and stiffness. This novel Berkeley disk resonator has a measured electromechanical coupling strength (Cx/Co) of 0.56% at 123 MHz without electrode-to-resonator gap scaling. This is an electromechanical coupling strength improvement of more than 5x compared with a conventional radial contour-mode disk at the same frequency. This increase should help improve the passbands of channel-select filters targeted for low power wireless transceivers and lower the power of MEMS-based oscillators.  

With the emergence of the Internet of Things in smart homes and buildings, determining the identity and mobility of people are key to realizing personalized, context-aware and location-based services - such as adjusting lights and temperature as well as setting preferences of electronic devices in the vicinity. Conventional electronic user identification approaches either require proactive cooperation by users or deployment of dedicated infrastructure. Consequently, existing approaches are intrusive, inconvenient, or expensive to ubiquitously implement. For example: biometric identification requires specific hardware and physical interaction; and vision-based (video) approaches need favorable lighting and introduce privacy issues. To address this situation, researchers at UC Berkeley developed an identification system that uses existing, pervasive WiFi infrastructure and users' WiFi-enabled devices. The innovative Berkeley technology cleverly leverages attributes such as the MAC address and RSS of users' WiFi-enabled devices. Furthermore, the Berkeley approach is facilitated by an unsupervised learning scheme that maps each user identification with associated WiFi-enabled devices. This technology could serve as a vital underpinning for practical personalized context-aware and location-based services in the era of the Internet of Things.

In our electronically connected society, human identification systems are critical to secure authentication, and also enabling for tailored services to individuals. Conventional human identification systems, such as biometric-based or vision-based approaches, require either the deployment of dedicated infrastructure, or the active cooperation of users to carry devices. Consequently, pervasive implementation of conventional human identification systems is expensive, inconvenient, or intrusive to privacy. Recently, WiFi infrastructure, and associated WiFi-enabled mobile and IoT devices have become ubiquitous, and correspondingly, have enabled many context-aware and location-based services. To address the challenges of human identification systems and take advantage of the popularity of WiFi, researchers at UC Berkeley developed a human identification system based on analyzing signals from existing WiFi-enabled devices. This novel device-free approach uses WiFi signal analysis to reveal the unique, fine-grained gait patterns of individuals as the "fingerprint" for human identification.

The emergence of huge, online digital repositories of data (AKA "big data") has made data mining challenging, especially for researchers, scientists, and businesses. These growing massive pools of online data have made it difficult to find relevant information, "connect the dots", and gain "big picture" perspective. For example, in the area of intellectual property, the access to global patent and trademark information includes billions of documents. To date, visualization of large-scale data sets is typically limited to two-dimensional tables, diagrams, and images. Many find these existing tools inadequate.To address this problem, researchers at UC Berkeley developed systems and methods for visualizing large amounts of data in three-dimensional virtual reality and augmented reality spaces. The initial application for this Berkeley technology has been patent documents. However, it's also applicable to visualize non-patent data, including technical and commercial data, etc.

Realizing the potential of massive sensor networks requires overcoming cost and power challenges. When sleep/wake strategies can adequately limit a network node's sensor and wireless power consumption, then the power limitation comes down to the real-time clock (RTC) that synchronizes sleep/wake cycles. With typical RTC battery consumption on the order of 1µW, a low-cost printed battery with perhaps 1J of energy would last about 11 days. However, if a clock could bleed only 10nW from this battery, then it would last 3 years. To attain such a clock, researchers at UC Berkeley developed a mechanical circuit that harnesses squegging to convert received RF energy (at -58dBm) into a local clock while consuming less than 17.5nW of local battery power. The Berkeley design dispenses with the conventional closed-loop positive feedback approach to realize an RCT (along with its associated power consumption) and removes the need for a sustaining amplifier altogether. 

Earthquakes are unpredictable disasters. Earthquake early warning (EEW) systems have the potential to mitigate this unpredictability by providing seconds to minutes of warning. This warning could enable people to move to safe zones, and machinery (such as mass transit trains) to be slowed or shutdown. The several EEW systems operating around the world use conventional seismic and geodetic network infrastructure – that only exist in a few nations. However, the proliferation of smartphones – which contain accelerometers that could potentially detect earthquakes – offers an opportunity to create EEW systems without the need to build expensive infrastructure. To take advantage of this smartphone opportunity, researchers at the University of California, Berkeley have developed a technology to allow earthquake alerts to be issued based on detecting earthquakes underway using the sensors in smartphones. Called MyShake, this EEW system has been shown to record magnitude 5 earthquakes at distances of 10 km or less. MyShake incorporates an on-phone detection capability to distinguish earthquakes from every-day shakes. The UC Berkeley technology also collects earthquake data at a central site where a network detection algorithm confirms that an earthquake is underway as well as estimates the location and magnitude in real-time. This information can then be used to issue an alert of forthcoming ground shaking. Additionally, the seismic waveforms recorded by MyShake could be used to deliver rapid microseism maps, study impacts on buildings, and possibly image shallow earth structure and earthquake rupture kinematics.

Energy Isolation Lock out Tag out (“LOTO”) is a series of CalOSHA and FedOSHA code compliance requirements and is the primary means by which equipment must be rendered “safe” prior to allowing personnel to work on the equipment.  LOTO codes require equipment-specific written procedures identifying all types of energy sources needed to operate the equipment as well as the energy-isolation methods and locations of utility disconnects, stored energy, etc. In addition, every LOTO procedure must be annually verified to confirm the written procedure is still accurate to the equipment.   Whereas current LOTO procedures are typically hand-written or using other time-consuming processes, UC Berkeley authors have created software allowing users to retrieve LOTO procedures in real-time guiding the end-user through a logical thought process to allow them to identify all energy sources and safety processes, and equipment needed.  

Trillions of sensors are envisioned to achieve the potential benefits of the internet of things.  Realizing this potential requires wireless sensors with low power requirements such that there might never be a need to replace a sensor’s power supply (e.g. battery) over the lifetime of that device.  The battery lifetime of wireless communications devices is often governed by power consumption used for transmitting, and therefore transmit power amplifiers used in these devises are important to their commercial success.  The efficiencies of these power amplifiers are set by the capabilities of the semiconductor transistor devices that drive them.  To achieve improved efficiencies, researchers at UC Berkeley have developed a novel method and structure for realizing a zero-quiescent power trigger sensor and transceiver based on a micromechanical resonant switch.  This sensor/transceiver is unique in its use of a resonant switch (“resoswitch”) to receive an input, amplify it, and finally deliver power to a load.  This novel technology also greatly improves short-range communication applications, like Bluetooth.  For example, with this technology, interference between Bluetooth devices would be eliminated.  Also, Miracast would work, despite the presence of interfering Bluetooth signals.

Information and communication technologies rely on integrated circuits (ICs) or “chips.” Increased integration has improved system performance and energy efficiency, and lowered the manufacturing cost per component. Moore’s Law predicts that the number of transistors on an IC will double every two years, yet industry experts predict that we are reaching economic limits of traditional circuit patterning processes. Photolithographic patterning is best suited to print linear features that are evenly spaced. The smaller or more complex the shape, the more likely the printed pattern will be blurred and unusable. Although multiple-patterning techniques can be used to increase feature density on ICs, they bring a high additional cost to the process. This means that the most advanced ICs available today have a high density of features, but are restricted to having simple patterns and are increasingly expensive to produce. Without innovations in production techniques, Moore’s Law will reach its end in the near future.  To address this issue, researchers at UC Berkeley have developed a one-step method to increase feature density on chips. This method is capable of achieving arbitrarily small feature size, and self-aligns to pre-existing features on the surface formed by other techniques. 

Background For high data rates and massive connectivity, the next generation cellular networks are expected to deploy many small base stations. While such dense deployment provides the benefit of bringing radio closer to end users, it also increases the amount of interference from neighboring cells. Consequently, smart management of interference is becoming a key enabling technology for high-spectral-efficiency, low-power, broad-coverage wireless communication. Over the past decades, several techniques at different protocol layers have been proposed to mitigate adverse effects of interference in wireless networks. One important conceptual technique at the physical layer is simultaneous decoding whereby each receiver decodes for the desired signal as well as part or whole of interference. When interference is strong this simultaneous decoding technique achieves the optimal performance for the two user Gaussian interference channel using good point-to-point codes. Moreover, it achieves the optimal maximum likelihood decoding performance in general, when the encoders are restricted to point-to-point random code ensembles. The celebrated Han-Kobayashi coding scheme, which achieves the best known performance for general two-user interference channels, also uses simultaneous decoding as a crucial component. As a main drawback, however, each receiver in simultaneous decoding has to employ some form of multiuser sequence detection, which usually requires high computational complexity to implement.   Technology Description Engineers from the University of California have developed a low-complexity coding foundation for communication channels with multiple pairs of senders and receivers, in which the signals from the senders interfere with each other and thus the signal observed at each receiver is a mix of the desired signal as well as one or more interfering signals and some noise. This technology will mitigate the adverse effect of interference caused by other communicating parties. More specifically, this technology decomposes a data stream into multiple substreams. These substreams are communicated over multiple units (“blocks”) of the span of time/frequency/space dimensions. Each sender encodes each of its substreams into a codeword that spans over multiple blocks and transmits multiple codewords simultaneously by superimposing them in a staggering manner. The characteristics of the codewords (coded modulation) and the mechanism of superimposing them (superposition) can be optimized with respect to the communication channel parameters as well as other transmission constraints. Each receiver recovers the codewords from its desired sender as well as some codewords from interfering senders by decoding its received signal over a sliding window of multiple blocks. For each window, multiple codewords (both desired and interfering) can be recovered one by one (successive cancellation decoding), which allows for each decoding step to be low-complexity. The selection of the codewords to be recovered as well as their decoding order can be optimized.

The increasing role of wireless technology is driving the need for reducing power consumption of wireless devices. The high-Q SAW and FBAR vibrating mechanical devices used for current RF band-pass filters are responsible for significant power savings. Still, there is room for improvement. To address this situation, researchers at UC Berkeley have developed an active resonator system with tunable quality factor, frequency, and impedance. Coupling two or more of these Berkeley resonators together enables construction of filters with arbitrarily small adjustable bandwidths and tunable insertion loss thereby achieving significant advantage over traditional filters constructed from passive resonators.

Frequency dividers have become essential components of phase-lock loops and frequency synthesizers that are used in a variety of applications from instrumentation to wireless handsets. In a typical frequency synthesizer application, frequency dividers often limit the achievable phase noise performance and contribute a large or even majority portion of the total power consumption. Common digital dividers offer good noise performance, but at the cost of power far in excess of that permissible for mobile applications and with poor scaling as frequency is increased. To alleviate this, injection-locked oscillator dividers have emerged as low power options at high frequencies, but they have performance limitations due to the active transistors used to sustain oscillation.To overcome these limitations, researchers at UC Berkeley have developed a new design for frequency dividers. While performing a frequency divide-by-two function, a version of this on-chip MEMS-based frequency divider reduced phase noise by 6 dB at close-to-carrier frequencies and 23 dB far-from-carrier. Unlike conventional frequency dividers, this Berkeley design dispenses with active devices and their associated noise, and operates with close to zero power consumption, limited in principle only by the power required to overcome MEMS resonator loss, estimated at 100 nW. With an output voltage swing of 450 mVpp generated from only 445 mVPP of input swing on a version of this MEMS divider, cascaded chains of fully passive dividers are possible, as needed for use in real-world phase-lock loops and frequency dividers. 

Data storage outsourcing has become one of the most popular applications of cloud computing, offering benefits such as economies of scale, flexible accessibility, efficiency, and allowing companies to focus on their primary business activities. Due to the increase in percentage of services conducted online and number of mobile internet connections, demand for data storage continues to grow. Customers in this industry are primarily concerned with authenticated storage and data retrievability. Although many efficient proof of retrievability technologies have been developed for static data, only two dynamic technologies exist. However, both are too expensive to implement in practice due to the fact that they require a high level of bandwidth. To address this problem, researchers have developed a dynamic proof of reliability scheme that requires 300 times less bandwidth than currently available technologies. This innovative technology makes dynamic proof retrievability of data practical and efficient, and thus attractive for the industry implementation. This technology gives clients of cloud storage providers assurance that their data has not been modified and that no data loss has occurred.

There is a long-standing problem of how to switch piezoelectric filters when used in switchable filter banks -- such as needed in RF channel-selection. To address this problem, researchers at UC Berkeley have developed a method and structure for a piezoelectric resonator with tunable transfer function -- i.e. tunable gain. This Berkeley resonator's gain is tunable to many values -- including values that are low enough to consider the device to be "off" relative to the background signal. Accordingly, this approach enables on/off switching of piezoelectric resonators; and it thereby obviates the need for separate low loss switches, which otherwise would be needed in series with piezoelectric resonators to switch them on and off -- adding insertion loss and raising system gain. In addition, this ability to adjust filter gain makes it possible for the resonator to control low power gain in a receiver front-end.