Danny Rittman, PhD
This paper describes an electronic tracking technology that is based on an advanced microchip system. This technology is implemented in a wide variety of ways, the most popular one is a “sticky patch” package called The Guardian Patch. The sticky patch can be affixed to any object in order to be tracked anywhere on Earth. The electronic circuit communicates with other, similar patches via a secured, encrypted private network. This system is called Guardian Tracking.
The electronic system includes a network connection for communicating with other, similar systems in a relay/mesh method through radio waves, computers, and mobile devices, including the Internet. The electronic circuit within the patch transmits signals to enable tracking anywhere on Earth – with or without GPS services. The system can store the affixed object’s details for identification purposes on mobile tracking apps. The system includes its own power source that lasts one (1) year. The system works in conjunction with a mobile software application to provide global tracking.
This paper describes the system’s components, functionalities, and method of operation from a high level perspective.
Guardian global tracking device – Introduction
Guardian Tracking is an electronic circuit that is a global tracking system in the form of a sticky patch. The sticky patch’s transceiver sends a beacon signal (PING) at defined intervals via a private, secured protocol in order to identify its exact location. The patch can be administered, managed, and controlled by a mobile or a web-based computer program through a secured, private network protocol.
Users can enter the affixed object’s details as a numeric/alphanumeric identification code. For example, for vehicle identification, the vehicle’s VIN number, license plate, owner information, or any other designated identification code, may be entered. This information is stored in the electronic circuit’s volatile memory.
The electronic circuit includes volatile memory, like flash memory, that has read/write functionality in order to write the object’s details. These details are written into the memory and will become the object’s identification data. These details are entered through mobile or permanent/web based software that works in conjunction with the patch.
Upon affixing the sticky patch on the object, the circuit is turned on. From then on, the electronic circuit will transmit an identification signal through GPS or short-wave at certain time periods. This will identify the device’s geographical location, worldwide.
The short-wave or GPS signal is received by a transceiver that is connected to a server with control software. The PATCH serial number is identified and the PATCH location is sent to the user’s mobile app via a conventional cellular network or through a website. Users can request the PATCH location at any time. A signal is then sent through GPS or short-wave to the patch. Upon receiving the signal, the PATCH acknowledges the request and sends its current location. The software assembles a visual map of the PATCH’s progress.
The system includes its own power source that lasts one (1) year.
Guardian app on an Apple Watch
Battery/POWER Harvesting and Management System
Energy harvesting is the process of capturing and accumulating byproduct energy as the energy becomes available, storing the energy for a period of time, and conditioning it into a form that can be used later, such as operating a microprocessor within its limits. Energy harvesting holds great promise for both low-voltage and low-power applications in a wide range of portable and mobile markets such as medical equipment, consumer devices, transportation, industrial controls, and military equipment. It is also a strong contender for use in applications that require a back-up battery, especially if the battery is in a remote or difficult to reach location. Energy harvesting will enable new applications and products that are presently impossible and even undreamed of.
Energy can be captured, or harvested, from sources deemed wasted or otherwise unusable. The process, also known as energy scavenging, captures residual energy as a byproduct of a natural environmental phenomenon or industrial process. It is therefore considered “free energy.” More often than not, this energy is simply released into the environment. Examples include mechanical energy resulting from vibration, stress, and strain; thermal energy from light bulbs, furnaces, combustion engines, and other heat sources. Other sources are energy from all forms of light; electromagnetic energy captured via inductors, coils, and transformers; wind and fluid energy from air and liquid flow; chemical energy from natural or biological processes; and huge amounts of RF energy in the environment from ubiquitous radio transmitters and television broadcasting.
In most cases, these sources provide energy in very small packets that have been previously difficult if not impossible to capture for later use. Energy-harvesting opportunities are enabled by new circuits that capture and store these small energy packets and channel them into useful purposes. The energy management provided by these circuits needs to include high-energy efficiency to capture and accumulate small energy packets, high energy retention to store the energy for long periods, and the proper energy conditioning to perform the desired task. The energy management must be well-defined
– and tolerate a wide range of voltage, current, and waveform inputs, including over-voltage, overcharge, and other irregular input conditions.
A variety of well-known devices, materials, or sensors are used to convert wasted energy into electrical voltages and currents, which can then be harvested, stored, and conditioned for many low-voltage wearable electronics and wireless sensor applications that previously required AC power or batteries.
Examples of energy generators include materials such as piezoelectric (PZT) crystals or fiber composites, solar photo voltaic cells, thermoelectric generators (TEGs), and electromagnetic inductor coils. These materials generate a wide range of output voltage and currents. None, however, can be utilized directly as power sources for driving low-energy electronics without energy-harvesting devices designed to capture the available power, manage it, and communicate handshake instructions to compatible wireless sensor systems.
In many cases, these sources provide energy as spurious, random, and otherwise irregular spikes, or in very low-levels. With recent developments in MOSFET “zero-threshold” transistor designs, energy- harvesting electronics have reached new heights, enabling capture, storage (in a capacitor, super- capacitor or battery), and management with high retention efficiency.
Efficient Energy – Capturing, accumulating, and storing small packets of electrical energy requires high efficiency. The circuit must stay in the active mode and be ready to capture harvestable energy. The device must be ready to provide an output as the application design requires it. For example, let’s say the energy is vibration from someone walking on a surface embedded with a vibration-energy source with a circuit, temperature sensor, and wireless transmitter. The small energy packets from infrequent pedestrian activity must power the circuit in the active mode for a long period until the circuit triggers the transmitter to send temperature data. Efficiency must be high; the energy consumed must be less than the energy from vibration.
Harvesting Systems – The classic (high-efficiency) energy-harvester consists of an energy generator, capture/storage/management electronics, and a load designed to be powered by the harvester – typically a wireless sensor network. In the block diagram below, a piezoelectric crystal membrane is shown as the energy generator. The piezoelectric generator transforms mechanical vibrations, strain, or stress into electrical current. This mechanical strain can come from many different sources, including human motion, other low-frequency seismic vibrations, aircraft or vessel vibrations, and acoustic noise.
Except in rare instances, the piezoelectric effect operates in alternating current, requiring time- varying inputs at mechanical resonance to be considered most efficient at generating energy. Most piezoelectric sources produce very high voltage but extremely low current, resulting in available power on the order of micro-watts – too small for most system applications, but an ideal source for energy- harvesting electronics.
AC energy from the PZT is channeled into the detector, which converts the voltage to DC and initiates the capture-and-storage operation. The detector can accept instantaneous input voltages ranging from 0.0V to +/-500V AC, and input currents from 200nA to 400mA, in either a steady stream of pulses or in an intermittent and irregular manner with varying source impedances. Early harvester electronics required a minimum of 4V input to capture and store the energy from PZT and other generators. More recent designs feature a front-end voltage booster and claim to initiate capture and energy storage with voltage inputs at less than 100mV.
As the source injects energy into the detector electronics, electrical impulses are collected, accumulated, and stored on an internal device such as a capacitor. The capture mechanism is set to operate between two supply voltage thresholds – +V_low DC and +V_high DC – corresponding to the
minimum (VL) and maximum (VH) supply voltage values for the intended “load” application. When (VH) is achieved, the output is switched to “on demand” to power the load. As the output diminishes and falls to (VL), the output is turned off and the charging cycle begins again until it reaches (VH). In one example, typical charge/cycle times are within four minutes, at an average input current of 10 uA, and within 40 minutes at an average input current of just 1.0 uA.
For optimal performance and energy retention, designing these energy-harvesting electronics must incorporate micro-power devices so that the energy consumed by the harvesting electronics is much smaller than the energy sent by the generating source. The net captured energy is a direct function of energy available for capture minus the energy the circuit must consume to stay in active mode.
Energy Preservation – A second key component of energy management is storage and retention with minimal leakage or loss. In the example of the bridge-monitoring application discussed above, when automobile traffic and vibration are minimized, there may be extended intervals before sufficient energy has been captured and stored. Therefore, the harvester’s electronic design must possess extremely high retention when the energy-generator function is randomly available or interrupted for long periods. Alternate energy generators such as solar or thermoelectric generators may also heighten available harvested energy.
Heuristic-based Power Management System
A heuristic (from the Greek for “find” or “discover”) is an experience-based technique that helps in problem solving, learning, and discovery.
In computer science, a heuristic is an algorithm which performs quickly and consistently, and/or provides good results. Hermes’s heuristic has a more specialized meaning: a dynamic set of algorithms, as opposed to a specific set of program instructions. A Hermes program doesn’t use just straightforward programming instructions. It has an advanced mechanism that no app currently has – a learning capability, which the computer world calls an “expert system.”
Guardian’s power management system is based on a smart heuristic. The electronics within Guardian work with its own rule-set, but are not limited to these rules. As the heuristic machine language software operates on Guardian, in conjunction with its smartphone app, it becomes key to prolonging the device’s battery life. Like the human brain, Guardian POWER Heuristic monitors physical conditions (temperature, humidity, altitude, etc.) and their effect on power consumption. Each of these factors is measured and “learned” in order to adjust the entire system power management, which conserves and maximizes power source life. The system target is to be fully operational for a period of at least one
(1) year, and as a living entity, it constantly learns about conditions and adjusts its power in order to continue operation. Furthermore, the system decides which sub-systems will be shut off for a period in order to conserve energy. The embedded heuristic is constantly learning and executing conclusions in order to operate longer. It can be compared to the human brain which is constantly expanding knowledge and computing power to perform complicated tasks – always.
Guardian has advanced POWER heuristic analyzer circuits that cycle through instructions before passing decisions to main sub-circuits. Utilizing the microprocessor for execution, they allow maximum power conservation. Guardian’s smart POWER Heuristic even puts itself to sleep occasionally to conserve power.
The most important aspect of power analysis validation is power consumption estimation and optimization. In order to handle these tasks the heuristic uses a range of search-tree-based analyses – an important field in theoretical computer science. The complexity of tasks in general is examined by studying the most relevant computational resources, such as execution time and space. Ranging problems that are solvable within a given amount of time and space into well-deﬁned classes is a very
intricate task, but it can help enormously to save time and to improve the overall system’s performance, which in our case is efficient power management.
Modern problems tend to be very intricate and closely related to analysis of data sets. Even if an exact algorithm can be developed, its time or space complexity may turn out to be unacceptable. However,
in practice the algorithm is often may be suﬃcient to ﬁnd an approximate or partial solution. This
extends the set of techniques to cope with the problem. As noted, heuristic algorithms suggest smart shortcuts and assumptions as viable solutions, exactly as the human mind does. Crucially, the heuristic has its own self-checking mechanism to verify that even if a shortcut decision is taken, it is viable and performing its task efficiently.
Guardian’s unique heuristic circuitry, or “wisdom,” combined with our mobile app, create an efficient POWER Management system to ensure long power life.
In this schematic, the RED sequences are active-power-consumption operations, the YELLOW are power-retention-mode operations (Listening), and the GREEN are sleeping. The POWER Management Heuristic ensures essential sub-circuit operations, listeners, and sleeping units according to physical conditions and operational necessities.
Guardian POWER Management silicon-based “wisdom”
Guardian’s operative, listening and sleeping sub-circuits – software heuristic
It’s diﬃcult to imagine the variety of existing computational tasks and the number of algorithms that have been developed to solve them, including heuristic algorithms. These algorithms include a broad spectrum of methods based on traditional techniques as well as speciﬁc ones. The simplest way to search algorithms is an exhaustive one that tries all possible solutions from a predetermined set, and then picks the best one.
For example, heuristics for the problem of intra-group replication for multimedia distribution services, based on peer-to-peer networks, are based on a “hill-climbing” strategy. “Divide and conquer” algorithms try to split a problem into smaller problems that are easier to solve. Solutions of the small
problems must be combinable to a solution for the large one. This technique is effective but its use is
limited because not many problems can be easily partitioned and recombined in such a way. “Branch- and-bound” technique is a critical enumeration of the search space that enumerates but constantly tries to rule out parts of the search space that cannot contain the best solution. “Dynamic programming” is an exhaustive search that avoids re-computation by storing the solutions of sub- problems. The attraction of this technique is formulating the solution process as a recursion. A popular method is the “greedy” technique that is based on the evident principle of taking the best choice at each stage of the algorithm in order to ﬁnd the global optimum of an objective function.
Heuristic algorithms are usually used for problems that cannot be easily solved. Time complexity classes are deﬁned to distinguish problems according to their difficulty. Class P consists of all those problems that can be solved on a deterministic Turing machine in polynomial time from the size of the input. Turing machines are an abstraction that are used to formalize the notion of algorithm and computational complexity. Class NP consists of all those problems whose solution can be found in polynomial time on a non-deterministic Turing machine.
Since Turing machines do not exist, their results mean that an exponential algorithm can be written for an NP-problem, though nothing is asserted regarding whether a polynomial algorithm exists or not. A subclass of NP, class NP-complete, includes problems such that a polynomial algorithm for one of them could be transformed to polynomial algorithms for solving all other NP problems. Finally, the class NP- hard is the class of problems that are NP-complete or harder. NP-hard problems have the same characteristics as NP-complete problems, but they do not necessarily belong to class NP. That is, class NP-hard includes also problems for which no algorithms at all can be provided. In order to justify application of some heuristic algorithm, we prove that the problem belongs to classes NP-complete or
NP-hard. Most likely there are no polynomial algorithms to solve such problems. Therefore, suﬃciently greater inputs heuristics are developed.
We designed Guardian with ambitious specifications. As the core program’s complexity puts it in an NP- complete category, the need for an out-of-the-box approach has come of age. We invented the “Learning Gopher” algorithm.
The Learning Gopher algorithm is not based on any known heuristic techniques. It introduces an entirely new heuristic approach and methodology. This is Guardian’s niche in the mobile application arena. It is patented in the US and internationally.
Let’s observe the state of the art heuristic techniques. Branch-and-bound and dynamic programming are quite effective, but their time-complexity is often too high and unacceptable for NP-complete
tasks. Hill-climbing is effective, but it has a signiﬁcant drawback called premature convergence. Since it is “greedy,” it always ﬁnds the nearest local optima of low quality. Modern heuristics try to
overcome this disadvantage. Simulated annealing algorithm, invented in 1983, uses an approach similar to hill-climbing, but occasionally accepts solutions that are worse than the current one. The probability of such acceptance is decreasing with time.
Tabu search extends the idea to avoid local optima by using memory structures. The problem with simulated annealing is that after overcoming obstacles, the algorithm can simply repeat its own track. Tabu search prohibits the repetition of recent moves. Swarm intelligence, introduced in 1989, is an artiﬁcial intelligence technique based on the study of collective behavior in decentralized, self- organized, systems. Two of the most successful types of this approach are Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO). In ACO, artiﬁcial ants build solutions by moving on the problem graph, and changing it in such a way that future ants can build better solutions. PSO deals with problems in which a best solution can be represented as a point or surface in an n-dimensional space. The main advantage of swarm intelligence techniques is that they are strikingly resistant to the local optima problem.
Evolutionary algorithms succeed in tackling premature convergence by considering a number of solutions simultaneously. Neural network searches are inspired by biological neuron systems. They consist of units called neurons and the interconnections between them. After special training on some given data sets, neural networks can make predictions for cases not in the training set. In practice,
neural networks do not always work well because they suffer greatly from problems of underﬁtting and
overﬁtting. These problems correlate with accuracy of prediction. If a network is not complex enough, it may simplify the laws which the data obey. If a network is too complex, it can take into account the noise that usually assists at the training data set while inferring the laws. The quality of prediction after training deteriorates in both cases.
The problem of premature convergence is also critical. Support vector machines (SVMs) extend the ideas of neural networks. They successfully overcome premature convergence since convex objective function is used. Therefore, only one optimum exists. The classical divide and conquer technique gives
elegant solutions for separable problems. In connection with SVMs which provide effective
classiﬁcation, it becomes an extremely powerful instrument. Later we discuss SVM classiﬁcation trees, whose applications present promising research.
Description and comparative analysis of simulated annealing, Tabu search, neural networks, and evolutionary algorithms provide successful, classical solutions in many software applications. But for Hermes we wanted higher standards.
Evolutionary algorithms are methods that exploit ideas of biological evolution, such as reproduction, mutation, and recombination for searching for the solution of an optimization problem. They apply the principle of “survival” on a set of potential solutions to produce gradual approximations of the optimum. A new set of approximations is created by the process of selecting individuals according to their objective function, which is called “ﬁtness” for evolutionary algorithms, and breeding them together using operators inspired by genetic processes. This process leads to the evolution of populations of individuals that are better suited to their environment than ancestors. The “Learning Gopher” includes a derivative version of a genetic algorithm as one of its core decision trees. In addition the “Learning Gopher” contains a combination of sub-heuristics to envelop the combinatorial explosion of possibilities with Guardian POWER Management topics.
The main approach of the Learning Gopher algorithm includes:
- Initialize and evaluate the initial conditions (population).
- Perform competitive, prioritized sub-circuits operations.
- Apply genetic operators to generate new directions, which will be later evaluated.
- Evaluate solutions in the population.
- Start again from point 2 and repeat until some convergence criteria are satisﬁed.
Sharing the common idea, evolutionary techniques can differ in the details of implementation and the
problems to which they are applied. Typical genetic programming searches for solutions in a specific domain. Their ﬁtness is determined by the ability to solve a computational problem.
Guardian’s Learning Gopher algorithm is different than typical evolutionary algorithms owing to its dynamic change of the analysis structure which allows a wide set of parameters to evolve. The Learning Gopher strategy works with real-world factors as representations of solutions, and uses self- adaptive mutation rates. The result is a self-contained POWER switching to the device’s sub-circuits in order to maximize battery life.
Learning Gopher algorithm includes one of the most successful evolutionary algorithms – a derived genetic algorithm. Genetic algorithms, first explored by John Holland in 1975, have demonstrated
essential effectiveness in many applications since then. GAs are based on the principle that the role of
mutation improves the individual quite seldom and, therefore, GAs rely mostly on applying recombination operators. They seek solutions of the problems in the form of strings of numbers, usually binary. The prevalent area for applying GAs are optimization problems requiring large-scale high- performance computing resources
One of Learning Gopher’s many objectives is to maximize the attainable capacity of packet-switching wireless networks. The main issue to consider for resource allocation is minimizing the number of Units of Bandwidth (UB) to be allocated. This type of problem has been proven in similar cases to be in the class NP-hard. Many applications applied GAs instead of greedy search that was used before. Here, we utilized the Learning Gopher for the device power analysis. It performs periodic condition simulation for idealized system power capacity utilization, causing substantial decrease of the system’s power consumption over time. The Learning Gopher includes an advanced technique that is domically modiﬁed according to various conditions in order to achieve smooth operation for at least one (1) year.
Our Learning Gopher approach within Guardian
Machine Learning and Pattern Matching mesh algorithm
Machine learning and pattern matching are powerful computer science techniques which can efficiently implemented within mesh/relay mathematical models. Guardian Tracking technology includes advanced mathematical model in order to autonomously control millions of “patches” systems. Since the Guardian Patch can be all around the world the system has to cover huge amount of data. The machine learning algorithm constantly gaining knowledge from the vast amount of data, and provides prediction and matching for each system. The algorithm controls the “sleep” and “listening” time of each patch in order to ensure area coverage at all time, worldwide. Since this type of information have extremely high complexity and gigantic data, it was developed as an adapting machine learning and pattern matching techniques including system physical veriﬁcation). Each system detection and data/pattern-driven physical location, determined by the level of abstraction and real time continuous simulations in order to provide good quality and results.
Machine learning is a computer science discipline that deals with the construction and study of algorithms that can learn from data. We applied machine learning techniques in our mesh based communication design to create what we call a supervised mesh learning, where the training data contains explicit self-learning case study of what the correct output should be for a given inputs. The system learns from each geographic situation and consisting of self-training and calibration in addition to self-testing/evaluation stages. In the training stage, given the input training features a set of
learning models are built. In the testing stage, the constructed learning models are to make predictions or decisions for each patch that is part of the mesh sleep/awake/listen state, rather than following only explicitly programmed instructions. We call this model Guardian Supervised Learning Algorithm (GSLA).
Our machine learning techniques are to assign a heuristic activation pattern to each one of the patches, while the pattern matching techniques are different in nature. Guardian pattern matching ﬂow is consisted of a set of patterns models that are constructed from the learning model. Different from a typical machine learning approaches, where usually there is no training stage to build up the library. Given a new testing/approve pattern, the system decides which pattern in library matches the current geographical condition and activates/deactivates patches in order to maintain efficient coverage at all times. Depending on whether the testing pattern is exactly matching to one pattern in the library, Guardian’s pattern matching is divided into many exact, individual pattern matching units and their matching conditions.
Control Logic System
Guardian control logic can be understood by using a state diagram, which is a form of hierarchical state machine. These state diagrams can also be combined with flow charts to provide a set of computational semantics for describing complex control logic. The control logic may be implemented using microcontroller or microprocessor. The control logic is implanted via hardware (IC) and a software program that together control the device’s operation. The control logic performs automated tasks that have been structured for the device’s functionality.
Guardian contains storage unit (memory). Storage technologies can be differentiated by evaluating certain core characteristics as well as measuring characteristics specific to a particular implementation. These core characteristics are volatility, mutability, accessibility, and addressability. For any particular implementation of any storage technology, the characteristics worth measuring are capacity and performance.
Non-volatile memory retains stored information even if not constantly supplied with electric power. It is suitable for long-term storage of information. Volatile memory requires constant power to maintain
stored information. The fastest memory technologies are volatile ones, although not always. Since primary storage must be very fast, it generally uses volatile memory.
Dynamic random-access memory is a form of volatile memory that also requires the stored information to be periodically reread and rewritten, or refreshed, otherwise it will vanish. Static random-access memory is a form of volatile memory similar to DRAM except that it never needs to be refreshed, as long as power is applied. It does lose its content when power is lost.
An uninterruptible power supply (UPS) can be used to give a computer a brief window of time to move information from primary volatile storage into non-volatile storage before the batteries are exhausted. Some systems, for example EMC Symmetrix, have integrated batteries that maintain volatile storage for several minutes.
Guardian contains both memories types (volatile and non-volatile) according to the system’s necessities.
Guardian memory system
The Affix Sensor System
Guardian’s affix sensor system is an advanced circuitry that controls the device’s mounting and removal responses. Guardian automatically turns on when affixed to an object. The “peel-and-stick” sensor system includes pressure and conductivity sensors that detect application to an object and start the device’s operation. It will stay on that object for the rest of the device’s life.
An internal clock starts to work. It has several tasks – one of them is determining the device’s lifetime.
If the device is removed before “end-of-life,” a sensor system turns on an emergency SOS transmission, alerting the user via the mobile app. The SOS transmits the device’s location until battery power is exhausted.
Guardian affix Sensors
GPS Receiver System
GPS receivers collect and process signals from the Global Positioning System (GPS) – a passive global satellite navigation system developed by the US Department of Defense. Depending on the quality of the receiver components, very accurate readings of time, position, and band velocity can be obtained. The same can be said of vector determinations between receivers. GPS is a satellite-based navigation system made up of a network of 24 satellites. GPS was originally intended for military uses, but in the 1980s, the government made the system available for civilian use. GPS works globally, in any weather conditions, 24 hours a day. There are no subscription fees or setup charges.
GPS satellites circle the earth twice a day in a very precise orbit and transmit signal information to earth. GPS receivers take this information and use trilateration to calculate the user’s exact location. Essentially, the GPS receiver compares the time a signal was transmitted by a satellite with the time it was received. The time difference tells the GPS receiver how far away the satellite is. Now, with distance measurements from a few more satellites, the receiver can determine the user’s position and display it on the unit’s electronic map. A GPS receiver must be locked on to the signal of at least 3 satellites to calculate a 2-D position (latitude and longitude) and track movement. With four or more satellites in view, the receiver can determine the user’s 3-D position (latitude, longitude, and altitude). Once the user’s position has been determined, the GPS unit can calculate other information, such as speed, bearing, track, trip distance, distance to destination, sunrise and sunset time and more.
ACCURACY – Today’s GPS receivers are extremely accurate, thanks to their parallel multi-channel design. Commercial GPS receivers have parallel channels and are capable of quick locks onto satellites and maintaining strong locks even in dense foliage or urban settings. Certain atmospheric factors and other sources of error can affect the accuracy of GPS receivers. Typically, commercial GPS receivers are accurate to within 15 meters.
Newer commercial GPS receivers with WAAS (Wide Area Augmentation System) capability typically improve accuracy to less than 3 meters. No additional equipment or fees are required to take advantage of WAAS. Users can also get better accuracy with Differential GPS (DGPS), which corrects GPS signals to within an average of 3 to 5 meters. The US Coast Guard operates the most common DGPS correction service. This system consists of a network of towers that receive GPS signals and send a
corrected signal by beacon transmitters. In order to get the corrected signal, users must have a differential beacon receiver and beacon antenna in addition to their GPS.
Differential GPS (DGPS) is highly accurate Image source: Garmin.com
The 24 satellites that make up the GPS space segment orbit the earth about 12,000 miles above us. They are constantly moving, making two complete orbits in less than 24 hours, and travel at roughly 7,000 miles an hour.
GPS satellites are powered by solar energy. They have backup batteries onboard to keep them running in the event of a solar eclipse. Small rocket boosters on each satellite keep them on the correct path.
There are 24 satellites orbiting the Earth at all times Image source: Garmin.com
GPS SIGNAL – GPS satellites transmit two low-power radio signals, designated L1 and L2. Civilian GPS uses the L1 frequency of 1575.42 MHz in the UHF band. The signals travel by line of sight, meaning they will pass through clouds, glass, and plastic, but not through most solid objects such as buildings and mountains.
SOURCE OF GPS ERRORS – A GPS signal contains 3 different bits of information – a pseudorandom code, ephemeris data, and almanac data. The pseudorandom code is simply a code that identifies which satellite is transmitting information. You can view this number on your commercial GPS systems, as it identifies which satellites it’s receiving.
Ephemeris data, which is constantly transmitted by each satellite, contains important information about the status of the satellite (healthy or unhealthy), current date, and time. This part of the signal is essential for determining position.
The almanac data tells the GPS receiver where each GPS satellite should be at any time throughout the day. Each satellite transmits almanac data showing the orbital information for that satellite and for every other satellite in the system.
Factors that can degrade the GPS signal and thus affect accuracy:
Ionosphere and troposphere delays – The satellite signal slows as it passes through the atmosphere. The GPS system uses a built-in model that calculates an average amount of delay to partially correct for this type of error.
Signal multipath – This occurs when the GPS signal is reflected off objects such as tall buildings or large rock surfaces before reaching the receiver. This increases the travel time of the signal, thereby causing errors.
Receiver clock errors – A receiver’s built-in clock is not as accurate as the atomic clocks onboard the GPS satellites. Therefore, it may have very slight timing errors.
Orbital errors – Also known as ephemeris errors, these are inaccuracies of the satellite’s reported location.
Number of satellites visible – The more satellites a GPS receiver can “see,” the better the accuracy. Buildings, terrain, electronic interference, or sometimes even dense foliage can block signal reception, causing position errors or possibly no position reading at all. GPS units typically will not work indoors, underwater, or underground.
Satellite geometry/shading – This refers to the relative position of the satellites at any given time. Ideal satellite geometry exists when the satellites are located at wide angles relative to each other. Poor geometry results when the satellites are located in a line or in a tight grouping.
Intentional degradation of the satellite signal – Selective Availability (SA) is an intentional degradation of the signal once imposed by the Department of Defense. SA was intended to prevent military adversaries from using the highly accurate GPS signals. The government turned off SA in May 2000, which significantly improved the accuracy of civilian GPS receivers.
Image source: Garmin.com
COMMERCIAL GPS RECEIVERS CHIPSETS – Today there are GPS Receiver’s IPs and complete chipsets/systems that can be purchased and used as plug-and-play units. The following are few examples of systems and manufacturers:
- FURUNO Multi-GNSS Receiver Chip eRide OPUS 7 ePV7010B – eRide OPUS 7 provides the world’s most-accurate positioning and navigation solution using simultaneous Multi-GNSS technology in combination with active anti-jamming, advanced multipath mitigation, and dead reckoning.
- MAXIM MAX2769 – The MAX2769 is the industry’s first global navigation satellite system (GNSS) receiver covering GPS, GLONASS, and Galileo navigation satellite systems on a single chip. This single-conversion, low-IF GNSS receiver is designed to provide high performance for a wide range of consumer applications, including mobile handsets.
Designed on Maxim’s advanced, low-power SiGe BiCMOS process technology, the MAX2769 offers the highest performance and integration at a low cost. Incorporated on the chip is the complete receiver chain, including a dual-input LNA and mixer, followed by the image-rejected filter, PGA, VCO, fractional-N frequency synthesizer, crystal oscillator, and a multibit ADC. The total cascaded noise figure of this receiver is as low as 1.4dB.
Maxim’s GPS IC MAX2769
- Digi-Key GPS Receiver – GPS technology is becoming more and more integrated with low-power modes. This mean tiny receivers can now be powered by a solar cell. One example is Retrievor, a collaboration of American, Australian, British, and Chinese companies that is using crowd sourcing to develop a coin-sized GPS tracking device. A small, self-powered GPS system can be used to track valuable items – even pets – using Android and Apple iOS apps.
The Retrievor unit measures 28 mm (1.10″) in diameter and 10 mm (0.39″) thick, integrating the antenna into the module to keep size down. It uses the SiRFstarIV GPS processor that enables operation in challenging GPS environments such as indoor tracking or when the end- user is on the move. This high level of GPS performance is achieved by using innovative firmware, which can detect changes in context, temperature, and satellite signals. It then updates its internal data whenever there is the opportunity for near-continuous navigation. Power for the Retrievor comes from an integrated solar panel and motion charger feeding a 3.7 V lithium-ion battery, which can also be charged via micro USB. User-defined ping rates can be adjusted from every second to once a day, so the Retrievor may never need recharging.
Guardian includes one of the top, off-the-shelf GPS Chipset units. It is controlled by the control logic unit and provides the geographical data that can be obtained when GPS services are available.
If GPS services are unavailable, the GPS receiver goes into sleep mode.
Radio waves are one form of electromagnetic radiation, as are microwaves, infrared radiation, X-rays, and gamma-rays. The best-known use of radio waves is for communication: television, cellphones, and radios all receive radio waves and convert them into mechanical vibrations in the speaker to create sound waves.
Electromagnetic radiation is transmitted in waves or particles at different wavelengths and frequencies. This broad range of wavelengths is known as the electromagnetic (EM) spectrum. The spectrum is generally divided into seven regions in order of decreasing wavelength and increasing energy and frequency. The common designations are radio waves, microwaves, infrared (IR), visible light, ultraviolet (UV), X-rays, and gamma-rays.
Radio waves have the longest wavelengths on the EM spectrum, ranging from about 1 millimeter (0.04 inches) to more than 100 kilometers (62 miles). They also have the lowest frequencies, from about 3,000 cycles per second or 3 kilohertz (kHz) up to about 300 billion hertz, or 300 gigahertz (GHz).
Scottish physicist James Clerk Maxwell, who developed a unified theory of electromagnetism in the 1870s, predicted the existence of radio waves. A few years later, Heinrich Hertz, a German physicist, applied Maxwell’s theories to the production and reception of radio waves. The unit of frequency of an EM wave – one cycle per second – is named the hertz, in his honor.
Hertz used a spark gap attached to an induction coil and a separate spark gap on a receiving antenna. When waves created by the sparks of the coil transmitter were picked up by the receiving antenna, sparks would jump its gap as well. Hertz showed in his experiments that these signals possessed all the properties of electromagnetic waves.
The National Telecommunications and Information Administration generally divides the radio spectrum into nine bands.
|Extremely Low Frequency (ELF)
|Very Low Frequency (VLF)
|3 to 30 kHz
|10 to 100 km
|Low Frequency (LF)
|30 to 300 kHz
|1 m to 10 km
|Medium Frequency (MF)
|300 kHz to 3 MHz
|100 m to 1 km
|High Frequency (HF)
|3 to 30 MHz
|10 to 100 m
|Very High Frequency (VHF)
|30 to 300 MHz
|1 to 10 m
|Ultra-High Frequency (UHF)
|300 MHz to 3 GHz
|10 cm to 1 m
|Super High Frequency (SHF)
|3 to 30 GHz
|1 to 1 cm
|Extremely High Frequency (EHF)
|30 to 300 GHz
|1 mm to 1 cm
Radio Waves Bands Image source: Livescience.com
According to the Stanford VLF Group, the most powerful natural source of ELF/VLF waves is lightning. Waves produced by lightning strikes can bounce back and forth between the Earth and the ionosphere, so they can travel around the world. Radio waves are also produced by artificial sources, including electrical generators, power lines, appliances, and radio transmitters. ELF radio is useful because of its long range and its ability to penetrate water and rock for communication with submarines and inside mines and caves. However, the carrier frequency is often lower than the frequency range of audible sound, which is considered to be 20 to 20,000 Hz. For this reason, ELF radio cannot be modulated fast enough to reproduce sound, which is why it is only used for digital data at a very slow rate.
LF and MF radio bands include marine and aviation radio, as well as commercial AM radio. Most radio in these bands uses amplitude modulation (AM) to impress an audible signal onto the radio carrier wave. The power, or amplitude, of the signal is varied, or modulated, at a rate corresponding to the frequencies of an audible signal such as voice or music. AM radio has a long range, particularly at night, but it is subject to interference that affects the sound quality. When a signal is partially blocked, the volume of the sound is reduced accordingly.
HF, VHF, and UHF bands include FM radio, broadcast television sound, public service radio, cellphones, and GPS. These bands typically use frequency modulation to impress an audio or data signal onto the carrier wave. In this scheme, the amplitude of the signal remains constant while the frequency is varied slightly higher or lower at a rate and magnitude corresponding to the audio or data signal. This results in better signal quality than AM, because environmental factors do not affect the frequency the way they affect amplitude, and the receiver ignores variations in amplitude as long as the signal remains above a minimum threshold.
Shortwave radio uses frequencies in the HF band, from about 1.7 MHz to 30 MHz, according to the National Association of Shortwave Broadcasters (NASB). Within that range, the shortwave spectrum is
divided into several segments, some of which are dedicated to regular broadcasting stations, such as the Voice of America, the British Broadcasting Corporation, and the Voice of Russia. Throughout the world, there are hundreds of shortwave stations, according to the NASB. About 25 privately-owned shortwave stations are licensed in the US by the FCC.
Shortwave stations can be heard for thousands of miles because signals bounce off the ionosphere and rebound back hundreds or thousands of miles from their point of origin.
FM and FM Stereo
As two-channel stereo music gained popularity, so did the demand for stereo radio broadcasting. However, one-channel (monaural, or mono) radios were already in wide use and were likely to remain so in the foreseeable future. The problem, then, was to create a system that could produce stereo music but still be compatible with existing mono receivers.
The method adopted for FM stereo broadcasting was rather ingenious. Ryan Giedd, a professor of physics at Missouri State University, explained that the broadcaster combines the left and right channels as L + R and L − R, and broadcasts them on slightly different frequencies, A and B. A mono receiver can lock onto A and hear both channels. A stereo receiver, however, locks onto both frequencies and combines A and B as A + B and A – B. A little algebra shows that A + B = (L + R) + (L − R)
= 2L, and A – B = (L + R) − (L − R) = 2R, thus effectively separating left and right channels.
SHF and EHF represent the highest frequencies in the radio band, and are sometimes considered to be part of the microwave band. Molecules in the air tend to absorb these frequencies, which limits their range and applications. However, their short wavelengths allow signals to be directed in narrow beams by parabolic dish antennas, so they can be effective for short-range high-bandwidth communications between fixed locations. SHF, which is affected less by the air than EHF, is used for short-range applications such as Wi-Fi, Bluetooth, and wireless USB. Also, SHF waves tend to bounce off of objects like cars, boats and aircraft, so this band is often used for radar.
Guardian includes a radio transceiver unit. A transceiver is a device comprising both a transmitter and a receiver which are combined and share common circuitry or a single housing. When no circuitry is common between transmit and receive functions, the device is a transmitter-receiver.
When GPS services are available, the device is obtaining its location and sends it through the radio transceiver. When GPS services are unavailable, the system is sending a radio wave with additional information for triangulation location-identification purposes.
Triangulation Direction and Source location Finding
Direction finding (DF), or radio direction finding (RDF), is the measurement of the direction from which a received signal was transmitted. This can refer to radio or other forms of wireless communication, including radar signals detection and monitoring (ELINT/ESM). By combining the direction information from two or more suitably spaced receivers (or a single mobile receiver), the source of a transmission may be located via triangulation. Radio direction finding is used in the navigation of ships and aircraft to locate emergency transmitters for search and rescue, for tracking wildlife, and to locate illegal or interfering transmitters. RDF was important in combating German threats during both the Battle of Britain and the Battle of the Atlantic. In the former, the Air Ministry also used RDF to locate its own fighter groups and vector them to German aircraft.
RDF systems can be used with any radio source, although very long wavelengths (low frequencies) require very large antennas, and are generally used only on ground-based systems. These wavelengths are nevertheless used for marine radio navigation as they can travel very long distances “over the horizon,” which is valuable for ships when the line-of-sight may be only a few kilometers. For aerial use, where the horizon may extend to hundreds of kilometers, higher frequencies can be used, allowing use of much smaller antennas.
Radio direction finders have evolved, following the development of new electronics. Early systems used mechanically-rotated antennas that compared signal strengths, and several electronic versions of the same concept followed. Modern systems use the comparison of phase or Doppler techniques which are generally simpler to automate. Early British radar sets were referred to as RDF, which as often stated was a deception. In fact, the Chain Home systems used large RDF receivers to determine directions.
Later radar systems generally used a single antenna for broadcast and reception, and determined direction from the direction the antenna was facing.
Guardian device works in conjunction with the static base unit hardware and software to accurately calculate the PATCH’s geographical location, without any GPS information. An advanced software that is executed within the static base unit analyzes the PATCH transmission and information. Based on triangulation and advanced mathematics calculations, computer software provides an exact location of the PATCH without GPS data.
AOA (Angle-of-arrive) is a method of getting the angle of a received signal from known stations to get user position. The angle of signal can be easily retrieved if the user device and beacon stations use directional antenna technology. However, the angle of stations might not always be the angle of received single because of multi-path.
Time-based triangulation uses distance to determine location. It assumes that the time used from beacon to user can be used to infer the distance between the two points, as the signal travels at near the speed of light. Time based-triangulation uses two methods: ToA (Time of Arrive) and TDoA (Time Difference of Arrive).
The ToA method directly measures the time a packet used to transmit from user device to beacon station – or vice versa. The user device can transmit a packet with a timestamp. The beacon can easily get the time of arrival, hence get the time for traveling. However, this method assumes that the time at the beacon station and user device is the same. To satisfy this assumption, the stations and user device must precisely synchronize their time, which is very hard to achieve in reality.
TDoA is similar to ToA, however it only requires beacons to synchronize their time. Even if the transmission time is unknown, the beacons can have different hyperbolic curves for the potential locations for different assumed transmission times. We can assume different hyperbolic curves for different assumed transmission times, and the curves’ intersection at a single point should be the correct transmission time. The point specifies the possible locations.
Beside the time, the property of a received signal is also an important way to infer distance. In most of the literature, RSS is used to represent received signal property. The propagation power-loss model has characterized fading signal strength over long distances. In reverse, the distance can be obtained by the strength of the received signal. In ZigBee network, LQI is regarded equivalent to RSS. LQI indicates the strength and data quality of link in IEEE 802.15.4. It is measured for every packet and is represented as an integer from 0 to 255.
The triangulation method has also been extended to multilateration in which there are more than three stations used to locate user position. Multilateration can combine any three stations to get the predicted result and use different measurements to weight the result and get the ultimate position.
When triangulation is considered, beacon stations and user need to be in line-of-sight. Otherwise, the angle or distance referred from time or RSS cannot used to locate the user. However, in a real world scenario, there might be walls, doors, rooms, and hallways in a building. Even if in a plaza, there might be furniture, statues, fountains, or pedestrians blocking line-of-sight.
The relay-based Guardian system
The Guardian device uses a few types of internal antennas. The device must use a GPS antenna and also an ELF one to ensure optimum performance.
A tiny portable device may be used in changing orientations, so an antenna element with a wide and uniform pattern may yield better overall performance than one with high gain and a correspondingly narrower beam. Conversely, an antenna mounted in a fixed and predictable manner may benefit from pattern and gain characteristics suited to that application. Evaluating multiple antenna solutions in real-world situations is a good way to assess which ones will best meet the needs of the application.
For GPS, the antenna should have good right-hand circular polarization characteristics (RHCP) to match the polarization of the GPS signals. Ceramic patches are the most commonly used type of antenna, but there are many different shapes, sizes, and styles available. Passive antennas are simply those tuned to the correct frequency, while active antennas add a Low Noise Amplifier (LNA) after the antenna and before the module to amplify weak GPS satellite signals. But they take more power than may be available from the energy harvesting source, as the VOUT line provides 2.85 V at 30 mA to power the
external LNA. The Guardian uses an embedded antenna type.
Embedded systems design is garnering a lot of attention. It is a fast changing and rapidly growing field. Antennas have always been one of the key components and challenges of RF designs, and probably will continue to be.
In the past, research concentrated on single bands moving higher in frequency. Now, by contrast, the push to high-frequency systems is taking place along with growing interest in low-frequency systems.
For designers of portable and embedded systems, the evolution from single band/narrowband to multiband/wideband has been rapid. Antenna designers report demand for penta-band coverage in the cellular bands (850 MHz to 2 GHz).
Complicating the drive to multiband systems is the desire for components to be as small as possible – and the antenna is no exception. The designers of antennas for wireless communication and radar systems have traditionally taken a conservative approach to the integration and co-design of the antenna with the radio frequency front-end circuitry. Usually, the antenna is designed separately from the electronics and manufactured in a different production technology. Standardized connectors and system impedances have normally been used to interface the antenna to radio electronics. However, as the operating frequency of the implemented systems increases, the complexity of the necessary packages and interconnections rises. The demand for compact, low-cost transceivers in short-range communication and sensor applications means that traditional design and assembly methods used in microwave and millimeter-wave technology have to be reconsidered.
Progress in the ﬁeld of circuit and systems design makes it feasible to ﬁt entire receivers and transmitters on a single semiconductor chip of less than 2mm, thus facilitating the design of highly compact systems utilizing several antenna elements for software beam-forming or increased system capacity. An increased number of antenna elements exacerbates the problem of ﬁnding compact, low- cost solutions for the implementation of antenna and high-frequency interconnects.
Integrated antennas offer a solution for the interconnect and antenna implementation problem by placing the antenna adjacent to the radio front-end, possibly on the same chip or printed circuit board as the electronic components. At millimeter-wave frequencies (>20 GHz), the size of simple low directivity radiators becomes comparable to typical integrated circuit sizes, thus enabling the use of on-chip antennas that completely bypass the interconnection problem, as illustrated in Fig. 1.1. One application for single chip transceivers is short-range communication in the license exempt ISM (Industrial Scientiﬁc Medical) 24.05-24.25 GHz and 60 GHz frequency bands. Car radar for anti-collision and intelligent cruise control systems are other mass-market applications that are planned in the 24 GHz and 77-80 GHz frequency ranges.
The Guardian device integrates several complex mixed-signal circuit blocks into a single system. One of the advanced features of the system is a constant self-check scan. The Guardian device is constantly searching for potential hardware, network, or software problems.
In case of a likely impending malfunction, the system first attempts to use alternate redundant units to ensure continued operation.
The Guardian device contains sub-unit redundancy. It has several blocks designed for self-correction and digital/analog control. The Guardian ensures smooth, continuous operation and in case of inevitable failure, it provides, via the mobile app, the estimated time until malfunction.
The Guardian device has constant self-testing capability and sub-system redundancy.
Working together (Mesh)
A mesh network is a network topology in which each node relays data for the network. All mesh nodes cooperate in the distribution of data in the network.
Mesh networks relay messages either by a flooding technique or a routing technique. With routing, the message is propagated along a path by hopping from node to node until it reaches its destination. To ensure all its paths’ availability, the network must allow for continuous connections and must reconfigure itself around broken paths, using self-healing algorithms such as Shortest Path Bridging.
Self-healing allows a routing-based network to operate when a node breaks down or when a connection becomes unreliable. As a result, the network is highly reliable, as there is often more than one path between a source and a destination. Although mostly used in wireless situations, this concept can also apply to wired networks and to software interaction.
Guardian mesh network is extraordinary. It’s controlled by a unique algorithm in order to manage “listening” time of the unit. This algorithm, which we call the “Listening Gopher,” is an evolutionary-
based algorithm that manages and controls each Guardian device’s “listening” time in order to participate in the mesh. Unlike a typical mesh where all components are constantly on, the Guardian mesh works with the GopherNET™ smart-timing protocol. This protocol ensures “wakeup” and “go-to- sleep” calls in order to efficiently manage system power and mesh operation.
Guardian devices work together via a private, secured communication protocol. Ensuring confidentiality and privacy, every Guardian device affixed on an object communicates with other units over a private network, creating enormous computing and database power worldwide. Once a Guardian device sends a transmission, the closest Guardian device receives it and passes it on to others, until the transmission reaches the main static transceiver. In this way, transmissions are passed on faster all the way to the main transceiver unit, and from there through conventional networks to the user’s mobile app.The combination of millions of units, world-wide, provides the best coverage and fastest response.
Guardian units work together, worldwide!
This paper has presented an overview of Guardian, a global tracking system. The system provides global tracking features, with or without conventional network/GPS. The system is small in size and operates on a power onboard source that lasts one year.
Key components include the antenna, radio transceiver, and power management system. Each sub- system is crucial for the entire system’s functionality. In addition an advanced software mathematical based algorithms are a key factor in the entire system’s functionality and global coverage. Using state of the art technology, we have achieved impressive results with prototypes. Further research will be conducted in order to improve the range, power management, and networking capabilities.
The Guardian team welcomes you to the future!
- A. Sangiovanni-Vincentelli, “Deﬁning platform-based design,” EE-Design, March 2002.
- R. Brayton, A. Hachtel, G.D., and Sangiovanni Vincentelli, “A survey of optimization techniques for integrated-circuit design,” Proceedings of the IEEE, vol. 69, pp. 1334–62, October 1981.
- S. Director, “The simplicial approximation approach to design centering,” IEEE Transactions on Circuits and Systems, vol. 24, pp. 363–72, July 1977.
- L. Carloni, F. De Bernardinis, A. Sangiovanni Vincentelli, and M. Sgroi, “The art and science of integrated systems design,” in ESSCIRC’02, September 2002.
- H. Liu, A. Singhee, R. Rutenbar, and L. R. Carley, “Remembrance of circuits past: Macro-modeling by data mining in large analog design spaces,” in Proceedings of DAC, 2002.
- R. Harjani and T. Tibshirani, “Feasibility and performance region modeling of analog and digital circuits,” in Analog Integrated Circuits and Signal Processing, Kluwer Academic Publishers, 1996.
- B. E. Boser, I. M. Guyon, and V. N. Vapnik, “A training algorithm for optimal margin classiﬁers,” in
5th Annual ACM Workshop on COLT (D. Haussler, ed.), (Pittsburgh, PA), pp. 144–152, ACM Press, 1992.
- B. Scholkopf, J. Platt, J. Shawe-Taylor, A. J. Smola, and R. C. Williamson, “Estimating the support of a high dimensional distribution,” Tech. Rep. MSR-TR-99-87, Microsoft Research, 1999.
- J. Platt, “Sequential minimal optimization: A fast algorithm for training support vector machines,” Tech. Rep. MSR-TR98-14, Microsoft Research, 1998.
- T. Joachims, “Making large-scale SVM learning practical,” in Advances in Kernel Methods – Support Vector Learning, MIT Press, 1998.
- C. Chang and C. Lin, “Training ν-support vector classiﬁers: Theory and algorithms,” Neural Computation, vol. 13, no. 9, pp. 2119–2147, 2001.
- S. Amari and S. Wu, “Improving support vector machine classiﬁers by modifying kernel functions,” Neural Networks, no. 12, pp. 783–789, 1999.
- S. Amari and S. Wu, “Conformal transformation of kernel functions: A data-dependent way to improve support vector machine classiﬁers,” Neural Processing Letters, no. 15, pp. 59–67, 2002.
- Christopher R. Houck, Jeffery A. Joine and Michael G. Kay, “A Genetic Algorithm for Function Optimization: A Matlab Implementation,” North Carolina State University, available at: http://www.ie.ncsu.edu/mirage/GAToolB.
- A. I. A. Cunha, M. C. Schneider, and C. Galup-Montoro, ”An MOS transistor model for analog circuit design. IEEE Journal of Solid-State Circuits,” 33(10):1510-1519, October 1998.
- P. E. Allen and D. R. Holberg, CMOS Analog Circuit Design, Oxford University Press, second edition, 2002.
- Gordon, T., and Bentley, P. “Towards development in evolvable hardware,” In Proc. The 2002 NASA/DoD Conference on Evolvable Hardware (15-18 July 2002), IEEE, pp. 241–250.
- Gruau, F. Neural Network Synthesis using Cellular Encoding and the Genetic Algorithm. PhD thesis, France, 1994.
- Gruau, F., Whitley, D., and Pyeatt, L. “A comparison between cellular encoding and direct encoding for genetic neural networks.” in Genetic Programming 1996: Proc. the First Annual Conference, MIT Press.
- Haddow, P. C., Hertmann, M., and Djupdal, A. “Addressing the metric challenge: Evolved versus traditional fault tolerant circuits.” in Second NASA/ESA Conference on Adaptive Hardware and Systems (AHS) (2007), pp. 431–438.
- Hartmann, M., and Goedeweeck, T. “Adapting a genotype-phenotype mapping to phenotypic complexity.” In NASA/ESA Conference on Adaptive Hardware and Systems (AHS) (2009), pp. 35–42.
- Hartmann, M., and Haddow, P. C. “Evolution of fault-tolerant and noise-robust digital designs.” IEE Proc. -Comput. Digit. Tech. 151, 4 (2004), 287–294.
- Hartmann, M., Haddow, P. C., and Lehre, P. K. “The genotypic complexity of evolved fault- tolerant and noise-robust circuits.” Bio systems 87, 2-3 (2007), 224– 232.
- Hartmann, M., Lehre, P.K., and Haddow, P.C. “Evolved Digital Circuits and Genome Complexity.” in Proceedings of 2005 NASA/DoD Conference on Evolvable Hardware (2005), IEEE Press, pp. 79–86.
- Kitano, H. “Designing neural networks using genetic algorithm with graph generation system.”
Complex Systems 4 (1990), 461–476.