US20180156618A1

US20180156618A1 - Devices and methods to facilitate escape from a venue with a sudden hazard

Info

Publication number: US20180156618A1
Application number: US15/885,906
Authority: US
Inventors: Thomas Danaher Harvey
Original assignee: Individual
Current assignee: Individual
Priority date: 2017-03-09
Filing date: 2018-02-01
Publication date: 2018-06-07
Also published as: US9921068B2; US20170176190A1

Abstract

A device and associated methods for escaping from a venue when a threat is detected is described. Venues can be buildings or outside areas and contain the area where the threat constitutes a hazard to a protected person. Threats include fire, terrorists, gunmen, explosion, collapse, loss of critical resources and crowd panic. The device incorporates a machine learning system implemented with a neural network or other pattern matching system and is trained in steps. Pre-training is based on general requirements such as edge-detection and audio analysis. Principles and data for venue layouts and human behavior can be included. The produced model is further trained from data gathered from sensors and servers after entry into the venue. Operation of the model produces warnings of threats and a plan of escape with steps of the plan communicated to the protected person by audio, visual or tactile sensory channels.

Description

CROSS REFERENCE TO RELATED APPLICATIONS

This application is a continuation of application Ser. No. 15/453,996 filed Mar. 9, 2017 now pending. Each patent application identified above is hereby incorporated by reference in its entirety.

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

Not Applicable

FIELD OF THE PRESENT INVENTION

The present invention is in the field of personal safety devices.

BACKGROUND CONCERNING THE NEED FOR THE CURRENT INVENTION

There are many devices available to provide direction and navigation assistance available in the current market place. But in emergency situations directions that are based on the normal situation of a location or venue are insufficient. Current information concerning the exact situation where a person is located, the nature of the emergency and all available clues for escaping to safety are desired. Historically, people have been willing to accept risks that would not be accepted today. Even though many hazardous situations and sudden dangers are relatively rare compared to daily events, many people are concerned about risks and want to have as much protection as possible.

BRIEF SUMMARY OF THE INVENTION

The current invention envisions devices and method using the devices to escape from a venue when a threat is detected. A processor constructs a model of a venue in its memory based on first information which is available prior to entry of a person to be protected from potential hazards. The model is trained by machine learning methods and receives further training after the person enters the venue. The model, thereby, takes into account the location of the person and information gathered by sensors about conditions on entry. Other pre-training can concern methods of escape from typical venues and behaviors by persons in the presence of hazards.
Observed conditions are evaluated by a processor in accordance with the model and the device generates an escape plan when a hazard is detected. The escape plan is displayed or transmitted to the protected person to facilitate escape from the threatened venue. The protected person thereby has an improved chance of successful escape in spite of difficulties that may appear from the threat or the conditions of escape. Warnings of a hazard may be detected and transmitted separately from the plan of escape.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING

The features and advantages of the various embodiments disclosed herein will be better understood with respect to the drawing in which:

FIG. 1 is plan diagram of a protected person entering a venue with an escape device.

FIG. 2 is plan diagram of a protected person leaving a venue with the guidance of an escape device after a hazard is detected.

FIG. 3 shows a front view of a typical embodiment of an escape guiding device

FIG. 4 shows a back view of the device of FIG. 3.

FIG. 5 shows a venue on board an aircraft.

FIG. 6 shows an escape device as part of the equipment of various persons.

FIG. 7 shows an embodiment of an escape device in a smart phone.

FIG. 8 shows an embodiment of an escape device in a vehicle.

FIG. 9 shows the steps and structure of the basic information flow of the device and its training.

FIG. 10 shows a diagram for process of developing software for an escape device.

DETAILED DESCRIPTION OF THE INVENTION

Escape Guiding Device
An escape guiding device, here usually called an escape device, is a portable device typically carried by a person entering a venue which collects information as the person enters and moves about the venue. If a threat occurs than the device provides guidance to allow the person to escape from the venue by a means or path that is calculated to allow the safest and quickest egress.
Typical Use in an Emergency Situation
A typical use is when a person has carried the device into a venue and allowed the device to continuously gather information which is used to train a computer learning system such as a neural net. When a threat such as a fire is detected by the device or the device is informed that a threat exist the device generates an escape plan with the model of the situation developed in the learning system. The steps of the plan are communicated to the person who uses them to escape.
Use by Persons with Limited Abilities
The escape device would be especially useful when carried by a person with special limitations on their abilities. For example, a person in a wheelchair who may or may not have the assistance of a second person to help move the wheelchair, could have the device especially developing an escape plan which is suitable for wheelchairs. This may take into account the restrictions or requirements on use of elevators during a fire or panic. Persons with sensory limitations such as blindness and deafness would also benefit from plans produced by the escape device which take into account their situation in relation to the situation in the venue.
The escape device when used by persons with limited abilities would in many embodiments have specialized display transducers. This would include audio outputs; speakers; vibrators; large, bright or flashing lights; and tactile devices. Another mode of use is by persons who use mobility assistance vehicles or modern motorized wheelchairs. The device can communicate with the chair at several levels from using resources of the chair as a display to complete autonomous control of the mobility device to carry the protected person from the endangered venue.
Limitation that affect egress may be sensory, cognitive or related to mobility. Similar considerations may apply to persons who do not have limited abilities but have responsibilities for other persons. This includes caretakers for elderly persons and for infants or small children. It would be necessary, for example, to make a special escape plan for a person who has a very wide baby stroller with triplet infants. Other factors would affect escape plans for devices where the protected person is a child. A child may not be able to compete with a panicked crowd of adults in pushing for a narrow exit.
Professional Use by Rescuers and First Responders
There are several professions where persons as part of their jobs enter into venues with high risks. These include firefighters and other first responders, the military in various operations, nuclear reactor emergency personnel and many other kinds. They can operate in area where conditions change quickly and information on current conditions may be hard to come by.
Venues
The escape device is envisioned as operating in an area called a venue in this specification and in the Claims. A venue here is an area which is sufficiently large to allow the assumption that if a person escapes from the area then that person is safe from an anticipated threat. The most common and typical venue is a building, but for some types of threats the venue may be limited to the portion of a building that accommodates a specific activity or some other area which contains hazardous zone. The extent of the venue is the area where the actual or anticipated threat is potentially operative.
Outdoor Venues
Some threats occur in an outdoor area. For example, a forest fire is a threat that is serious to people in certain situations such as to people fighting that fire. The venue in this case is the area that is threatened by the fire. Other outdoor venues include dangerous neighborhoods, battlefields, and flood plains. The venue is defined by the range and nature of the threat. Another example is in a situation where a car has broken down in a limited access highway intersection. This is an area dangerous to pedestrians and an the venue extends until an area safe for pedestrians is reached.
Special Venues
In one series of embodiments an escape device can be used to escape from the cabin of a passenger aircraft after that aircraft has crashed. On entry to the aircraft the device determines the seat that the protected person is occupying and consults a contained or downloaded database of aircraft layouts to determine the best escape path. Accelerometer and gyroscope sensors can evaluate the nature of the crash and with clues from a camera a reasonable escape path and procedure can be generated.
Another important type of venue is an area where industrial or other technical operations is being performed. When things are going wrong escape from a chemical plant, a reactor site, a ship, or a crime scene may be in order. An escape guiding device is useful in such situation, especially if the protected person is not familiar with the specific location or type of venue.
Threats.
The types of threats that may occasion the need to escape from a venue vary with the nature of the venue; but there are many different kinds in each class of venue that may require such action.
The most well-known threat is fire. Fire can occur in buildings and trap people by blocking exits and creating situations where exit by the standard route or the entry route is impossible or inadvisable. Fire creates a lot of physical clues such as sounds, heat and smoke; but it can also be hidden. Subtle use of sensors and information from servers may be necessary to assess the extent of danger and to create a workable plan to escape the threat or the venue.
With respect to buildings, explosion and collapse are related to fire because the original cause may be a fire. These events are usually sudden, and an escape device is faced with the problem of evaluating data to determine what happened, where it happened and what limitations have been created for an escape plan. In such situations, it is useful to have access to data from servers concerning the details of the venue. Such data may be currently accessed or accessed prior to entry to the venue. It may be used to pre-train the computer learning model allowing additional training from new data concerning the threat event to allow generation of an escape plan.
Wildfire which is commonly referred to as forest fire in certain cases also provides a threat that requires a sophisticated escape. There have been several cases where firefighters in rugged terrain have been suddenly overtaken by flames sometimes with tragic outcomes. In many cases, local sensors would be unable to detect the threat coming; but with server input concerning the threat a computer learning system can combine that information with local sensor data concerning current location and conditions observed in the venue to generate a reasonable path. Such escapes sometimes need to be very quick and continuous computer monitoring of the situation with continuous training of a model would often be needed.
Terrorism and the related possibility of attach by hostile persons for other reasons has become a wide concern. Local sensor data concerning such events as sounds of gunfire, crowd noises and audible instructions from venue staff or other persons can be combined with venue layout and location information in forming a plan.
The listed threats, other causes, and events caused by the dynamics of otherwise unthreatening situations can cause panic. Panic itself can be a danger and threat that exceeds the importance of its original cause. Stampeding crowds have caused many injuries and fatalities. An escape device should in its basic levels of training take into account these possibilities and design escape plans which avoid particular crowd dangers. These include narrow restrictions and cul-de-sacs which threaten a protected person with being suffocated or crushed by crowd pressure.
The threats listed above contain obvious serious dangers to persons; but there are other threats which seem less serious but in certain cases may be important. Nursing home residents, prisoners and other persons with limited abilities to help themselves can be seriously threatened with loss of resources. These threats include loss of power and light, being locked in, loss of elevator service as well as other problems. Even when staff or other personnel are available to help with the situation, the nature of the difficulty or the number of persons threatened may be overwhelming. For example, elevators and building lighting are in most cases backed up by local generators in the event of a power supply outage. Unfortunately, generators often disclose their inoperability only after they are needed. An escape from a high floor in an unlighted stairwell is very difficult. Even if most threatened persons can “shelter in place” it may be necessary to send someone for help or needed supplies.
Escape.
Escape from a venue under threat means to move away from the venue to a place where the threat is no longer a danger. Escape from a threat in a venue means to either move away from the operative area of the threat or to take action which renders the threat inoperative. The person carrying the escape device could be a leader of a group and could be charged with leading the entire group to safety.
A key function of an escape device is to form an escape plan. An escape plan can include a path to travel, conditions to be check for which would modify the plan, and conditional actions in the event of determination of necessary conditions. An example plan could be expressed in words as “travel back to the entrance, but if it's blocked by a crowd go to the side door.” The sensors in the device inform the model which would in effect watch for signs of a crowd and redirect movement to the side door if they are found. The plan may contain routes or paths to escape from the specific venue which comprise locations and directions to facilitate movement away from the hazard or venue.
While the entire plan could be made available to a person using the device, in most cases the device would display or convey to the user the next step to take in escape. Keeping the presentation simple may be very important in a sudden threat situation.
Sensors
The escape guiding device can employ information from a variety of types of sensors. These include video or picture cameras, accelerometers and gyroscopes, GPS receivers, receivers for transmissions from remote sensors, threat sensors not a part of the device. The term sensor as used in this specification and in the claims, includes not only sensors that are part of the device but also receivers that collect currently sensed information and are provided by the venue or third parties. It also includes a GPS receiver where the receiver generates location information from timing details of the received GPS signals.
Cameras
A camera either still or video can gather information. Machine vision software is now available which will recognize situations and objects and can provide a critical input to a computer learning system. Such learning systems often work in several layers and provide learning at a high level in a manner using generic detail recognizing systems at a lower level.
Accelerometers and Gyros can be used to detect movements of the device or the person carrying the device. In the phase prior to detection of a threat, such information can be correlated with venue layout, GPS and other information to give a more detailed picture of the entry path. Slowdowns, halts, and diversions on entry may or may not be clues to potential difficulties on egress. That may be determined by correlation with other information. In the escape phase after a threat is detected other movement events are important information. Slowdown, halts and diversions as even more important here. Other events such as a person falling or being knocked down can be detected and taken into account. The rate of movement and details such whether the movement of the person escaping are important
Microphones
Microphones can be used to detect conditions around the escape guiding device in several ways. If multiple microphones are in use and have directional sensitivity that can be correlated with directional information from sensors.
One category of useful sounds are those made by other persons in the venue can give important clues to the movements of persons, to the number of persons in various area and the experiences that persons are having. It may be possible by using analysis of the sounds of persons to detect potential escape routes which are blocked or where panics are occurring.
Another category of useful sounds are those made by the threat itself. Fires and gunshots as well as the panicked crowds mentioned above have characteristic sounds which are subject to analysis.
A third category of useful sounds are announcements and other information given by the venue operators. Sensitive microphones and analytic software can interpret these sounds where they would be covered up by other ambient sounds as they are heard by persons.
Miscellaneous Sensors
Other types of sensors which may be useful include magnetometers which can give an absolute direction from the earth's magnetic field in certain cases and thermometers which can detect heat from threats and changes of temperature from outside area in cold weather.
Output Device.
The term “output device” is used in this specification and in the claims rather than simply “display” to convey that the output of warning and guidance for the protected person may be in modes other than the common visual screen. Especially with persons of limited abilities and for most persons in difficult environments, a visual screen may not be the best way to convey the needed information in a way to get timely escape action. Some output devices work by conveying information to a person and some by conveying information to equipment such as an autonomous vehicle which acts on an escape plan.
Some output devices are visual. They can be general display screens which can show pictures and text and provide detailed instructions. These can be integrated with input devices such as touch screens. Other visual output devices include lights which can be flashing to get attention and mechanical devices which raise flags.
Some output devices are audible. This includes speakers and earphones as well as sirens which may be operated over a communication link.
Tactile output devices include vibrators, braille devices and specialized devices which operate mechanical signifiers.
The escape device can operate vehicles as the primary or an additional output mode. In other cases, the escape device can operate elevators, doors, open gates and substitute for a key or pass code to allow exit.
Beacons
Beacons can be placed in the device by the venue operator. The beacons can operate by radio, infrared or other communication means. An ordinary Exit sign is in effect a beacon. Such signs may incorporate beacons that operate in other modes for use with escape devices. More sophisticated venues may come to offer services with venue information at their locations. In the future, digital escape information may come to be required as an extension of the services for direct communication to humans. Current flashing fire warnings, sirens, buzzers, exit signs etc. may have digital equivalents and augmentation. An intelligent escape device would take advantage of such information as much as possible.
Processor, Machine Learning and Models
A device to use gathered information to guide a person from a venue where a serious hazard exists requires very sophisticated computer system to complex problems in interpreting incoming data and generating a plan for escape. In general, it is not feasible to discover all of the rules and relationships necessary to solve that problem and to write a determinative computer program that produces a sufficient result. However, methods have been developed and are wide and increasing to use a set of examples which is processed and used to product a set of rules which working together can find answers suggested by the examples. A substantial set of examples and a large amount of processing are required but many people are trained and are being trained in application of well known methods to implement this approach on a wide variety of problems.
There are limitations on the kind of problems that can be solved with this approach, but the problem here is well suited for the approach because of the specific input data that is to be gathered and because of the specific kind of output that is required.
Machine learning as used here and in the Claims is a term for the type of artificial intelligence that is implemented without explicit programming to generate solutions to the problems confronted. It is focused on the development of solutions by having the effective program change as it incorporates data in such a way that it converges on an ability to produce the required solution.
Model
The term model as used in this specification includes representations that do not explicitly describe the system modeled but which are designed or trained to produce information based on the behavior of the system. It includes pattern matching system such as a neural network which has been trained on data exemplifying the system. In that case the model consists of a, probably huge, array of parameters that determine the operation of the individual neurons in the neural net program. Training would work by systematically adjusting the values of these parameters on the basis of the training data.

DETAILED DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING

Referring to FIG. 1:
A person 10 to be protected is entering into a venue at an entrance 11 carrying an escape device 12. The device gathers information as the venue is entered and the person progresses through the venue along path 13. The device gathers information from sensors which may include an optical camera(s) with a field of view 14 and self contained sensors such as an accelerometer or gyroscope. A beacon(s) 15 in the venue may provide additional information to the device with a suitable signal 16. The device both before and after entering the venue may receive signals 17 from the outside servers which provide various kinds of information such as location (gps), venue layout, pre-calculated paths and parts of paths and many other kinds of useful data. All or portions of the calculation burden may be offloaded to external servers. The protected person proceed along the path which may contain various possible areas 18 where on return false turns may be taken or additional dangers may exist. The inner parts of the venue are the destination and the activities for which the person has come take place 19.
Referring to FIG. 2:
The person 10 from FIG. 1 is in the inner part of the venue with the escape device 12 when a threat 20 appears. The threat depicted is a fire which has broken out. A large number of other people have entered the venue. If the venue is full of panicked people as depicted (numbers are omitted for clarity) and the lighting has failed, efficient escape may be difficult. In many cases the best escape is not back through the entrance path, but in the depicted case that path has been chosen by the device and is shown as 13. Two cull-de-sac's 18 are shown. They may not have been noticed by the protected person on the way in; but they are each holding a crowd of people who think that is the way out. Without the help of the escape device, the protected person may choose to join the people trying unsuccessfully to force an impossible exit.
Referring to FIG. 3:
A front view of one embodiment of an escape device is shown. The depicted device is designed to be worn on a chain, ribbon or lanyard around the neck of a protected person with a provided eye 30 for attachment. The central area 31 of the device is a display screen which gives instructions in case of a detected hazard and shows a decorative screensaver chosen by the user at other times. A camera 32 is used to gather information on entry to the venue as the device hangs around the protected person's neck. Eight sound ports 33 lead to microphones which gather audio information to be analyzed for content, timing and arrival direction. A fringe 34 around the device is for decoration and to help distinguish the two sides of the device. As depicted the device is showing the detection of a threat 35 and is beginning the process of guiding the protected person from the venue with exit directional information 36. The screensaver decorative display has been replaced with a warning.
Referring to FIG. 4:
The back of the device is shown. It is cut away as necessary to show interior components. A speaker 40 is included to alert the user. Various input and sensors are shown including an accelerometer 41, a gyroscope 42, a GPS receiver 43 and a communication device 44. The communication device can be used for many purposes such as receiving beacons from the venue, information about the venue from an outside server, access to outside processing assistance from a server or download of pre-trained parts of a pattern recognition or neural network for the device. Use of the outside servers allows function with a reasonable processing load on the devices own processor 45.
Referring to FIG. 5:
A venue is depicted where the escape device is adapted for escape from an aircraft cabin 50. The protected person enters from the normal cabin door 51 along path 52. The protected person arrives at assigned seat 53. The device integrates information downloaded from servers such as the persons assigned seat, the layout of the particular aircraft for this flight, the anticipated number of passengers and the scheduled times of the flight. This information is used in combination with information gathered by the sensors as the protected person enters the aircraft and after a hazard is detected. If a hazard situation is detected the escape device evaluates all available information and plots an escape route. In the instant situation, the escape route 54 is not toward the entrance door but is in the other direction to the exit aisle 56 which leads to an emergency exit 56.
Referring to FIG. 6:
The inclusion of an escape device in the equipment of persons in various roles is shown. The device is designed to be able to access servers and the environment with sensors and communication device, perform it's analytical role and provide escape information to the user. It may also be integrated into other equipment associated with the protected persons role both for purposes of gathering information and for outputting escape paths.
One situation where integrating an escape device into a person's daily equipment is in the case of a firefighter 60. The device could have sensors 61 and a display and warning device 62. In this case it would be helpful to provide specialized sensors such as infrared heat detectors. If the firefighter comes to a closed door with intense fire on the other side, it is dangerous to open that door. Firefighters are so trained, but forgetting this critical rule is a source of many injuries. If the firefighter is attempting to escape other critical dangers, the chance of making such a mistake is multiplied. Heat sensors in the escape device could warn the firefighter not to take that route and the device could suggest another escape plan.
Another situation is that of a military person 63 or soldier. Many battlefield conditions can affect the choice of a suitable escape route from a dangerous area. Directional sound detectors in a sensor module 64 would be useful in detecting area with friendly or hostile occupants or active use of weapons. Taking these factors into account a computer learning system would be able to estimate an appropriate route for accessing a safer area such as in returning to ones unit. Directions could be displayed in a small display unit using an appropriate sensory mode 65.
A forest firefighter is shown at 66. In this case the protected person is using a water pack to put out a small fire separated from the main area of threat. Heat sensors as part of a sensor package 67 would be useful to detect the approach of active burning and may save a person in danger of being surprised. In particular, continuous planning of an escape route can cause a danger of being cutoff to be detected before the actual threat approaches. Again, a display and warning device is used 68.
A mobility chair or motorized wheelchair is shown at 69. The sensor unit 70 gathers information from servers and sensors. In the event of a threat after a plan is developed appropriate instructions are display on a display 71. In some cases outputs from the escape device are used to direct the chair controls to make the escape or facilitate the needed movements. For example, the chair may have built in limitations of speed and other parameters that can be overridden.
Referring to FIG. 7:
An embodiment of an escape device in a smart phone 80 is shown. Typical smartphones have many relevant sensors 81, substantial processing power and visual, audible and vibratory output devices 82. With suitable software embodying a learning model of venues an escape device can be implemented as an application in a smartphone.
Referring to FIG. 8:
An embodiment of an escape device integrated into a vehicle 90 is shown. The escape device uses a sensor and communications module 91 mounted on top of the vehicle. When a threat is detected the plan developed by the escape device is displayed on the GPS display 92 integrated into the vehicle. The input devices for the vehicle GPS 93 are also used to enter parameters for the escape module. In this way the vehicle user can instruct the escape device about servers and sources of information to be considered.
Referring to FIG. 9:
The steps and structure of the basic information flow of the device and its training are shown. There is a timeline 100 which is divided into three periods. First pre-entry 101 which is the time prior to use of the escape device to track entry and movements into a venue. The device is pre-trained during this time. The next time period is pre-threat 102, the time prior to detection of a threat which is used to train the escape device and store relevant conclusions. The final time period is the threat and escape period 103. During this time collection of data and training continue but evaluation of the model and development of escape plans and instructions also occurs. The instructions are put to the output display transducers to allow the protected person to escape the venue.
During the pre-entry period, the model comprised in the escape device is constructed and initialized to parameter values suitable to allow effective training 104. Pre-training 105 typically proceeds in two stages and uses two data sets. The first stage uses a generic data set and trains for general skills such as recognizing objects and edges from a camera sensor. The data set for this stage may not be one specifically developed for this application but is suitable to the sensors and processing resources to be used in the device. The second stage used a data set developed 105 specifically for this application which is based on real or virtual escape and venue analysis scenarios. Such data sets would find data concerning specific venue layouts and facilities and data concerning typical venues that can predict possible situations in actual venues to be especially relevant. The data sets are trained 106, usually sequentially, into the model. Low level training. such as here used, is computation intensive; but it can be accomplished on a one-time basis using substantial computing resources not needed by the final escape device. If the device is produced in mass quantities the pretraining results are loaded into each device in the process of production. Additional data sets and training can be done prior to use of the device to adapt it to specific anticipated uses or venues.
During the pre-threat period, the protected person takes the escape device into the venue. Sensors in the device collect 107 information about the venue and the protected person's location and progress. Information is also downloaded 108 from various servers and beacons provided by the venue or outside sources. The sensor and server information is used to do more venue specific training of the model. This can be accomplished by a processor in the escape device; but it may also be assisted by training resources located in servers away from the escape device. Sensor and server data can also be stored 110 to be used as an input in the later threat stage. The escape device is then readied to receive a threat.
During the threat period, the escape device has been told or has determined that a threat is detected and escape plans and actions should be undertaken. The model continues to be trained 111 from all information sources including stored data 112 and data that continues to be collected from sensors 113 and servers 114. The model is evaluated 115 based on its current training 111, stored data 112, continuing sensor data 113 and continuing server data 114. Outputs to guide the protected person are generated 116 and presented for action.
Referring to FIG. 10:
A diagram of the process for developing the pattern matching software for the embodiment of FIG. 9 is shown. The software to be developed is a recurrent neural net with an input layer, a recurrent layer, a additional fully connected hidden layer and an output layer. Other more complex structures are also easily implemented but this structure would be sufficient for the low level of accuracy needed in this embodiment.
The first step of the development is to accumulate 120 a data set for training and testing. Several data sets are required as described for FIG. 9. One set of data is collected by a device that is constructed similarly to that of FIG. 3 and the processor is programmed to collect data from the sensors and transmit it via the communication module to an outside data collection system. A substantial number of escape scenarios enacted or simulated and the resulting data is manually graded and entered into the database.
The data is divided into two sets with a larger portion for training and a smaller portion for accuracy testing. This is considered labeled data because it contains both input (sensor) data and the desired output for that data (presence or absence of transfer).
In this and related embodiments, a step in the development which might be started in parallel with data collection is the design of an appropriate neural network. The sizing of the layers and the setting of various factors in the neural net which are in addition to the factors and values (parameters) that are adjusted in training are collectively referred to as hyperparameters to distinguish them from the “parameters” which are adjusted in training the neural network. The hyperparameters are initialized 121 to appropriate values. In some systems that are taught hyperparameters are adjusted during the course of training but are distinct from trainable parameters because the adjustments are on the basis of the progress of the training rather than being direct functions of the data.
The next step is to initialize 122 the parameters which are to be trained. Appropriate initialization is necessary for reasonably rapid convergence of the neural net. A number of techniques are taught to product an initial set of values which produced good training progress.
The network is then trained 123 by passing data set items through the network as implemented on a training processor. Because training requires larger processing power and time than use of the network after training special powerfull processors are used for this step. The training process adjusts the parameters incrementally on the basis of the output of the neural network. The hyperparameters specify the methods of calculating the adjustment to parameters. Generally, the output of the network is used to back propagate through the network to provide further input to the adjustments. The items in the training portion of the dataset are used repeatedly while the convergence of the network is observed 124 by processes in the training data processor.
If the convergence is judged 125 not to be adequate the training is stopped, the hyperparameters are adjusted 126, the neural network is reinitialized and the training process is repeated until satisfactory convergence is obtained. The smaller portion of the data set which has been retained and not used for training is then passed 127 through the neural network (classified) and the output is checked 128 for accuracy. If accuracy is not sufficient for the goals of the particular system being developed then the net structure is made larger 129 and the training process is repeated until satisfactory accuracy is obtained.
The trained neural network is then downloaded 130 to the target device, which is then ready for system testing 131.

Claims

I claim:

1. A device to facilitate movement in a venue comprising:

(a) a first processor with a memory containing a program for machine learning and a model of a venue wherein the model has been pre-trained using a second processor with information concerning at least one of layout of venues, methods of escape from venues, and behavior of persons attempting to escape from a venue in a hazardous condition;

(b) a sensor to detect a person in the venue and a sensor to acquire information to be used in training the model using the first processor concerning at least one of a location of the person in the venue and behavior of persons attempting to escape from the venue;

(c) a program comprised in the memory of the first processor to use the model and data from at least one of the sensors to produce an instruction to escape from the venue subsequent to the detection of a hazard in the venue; and

(d) an output device using the instruction to guide the person in escape from the venue.

2. The device of claim 1 wherein:

the instruction to escape from the venue is determined by training of the model with data acquired from the sensor subsequent to entry of the person into the venue.

3. The device of claim 1 wherein:

the instruction to escape from the venue is determined by training of the model from information acquired from a server subsequent to entry of the person into the venue.

4. The device of claim 3 wherein:

the information from the server is selected on the basis of the information acquired from the sensor.

5. The device of claim 1 wherein:

the output device is a least one of a visual display, an audible transducer and a tactile transducer.

6. The device of claim 1 further comprising:

a vehicle to escape the venue wherein the vehicle is adapted to be operated autonomously to escape from the venue in accordance with the instruction.

7. The device of claim 1 wherein:

the output device comprises a vehicle to carry the person.

8. A device for facilitating escape from a venue comprising:

(a) a sensor adapted to be carried by a person entering the venue and to acquire information about potential escape routes from the venue,

(b) a memory containing at least one of prerecorded information concerning physical layout of the venue and information acquired from a communication system concerning physical layout of the venue,

(c) a first processor with a program to calculate an escape route based on the memory and the information gathered by the sensor, and

(d) an output device to guide the person escaping from the venue when a threat is detected by the device comprising at least one of a visual display and of an audible transducer;

wherein:

(a) the program comprises a machine learning model transferred from a second processor subsequent to training concerning methods of escape from venues in hazardous conditions; and

(b) the program has been trained using the information.

9. The device of claim 8 wherein:

the program is configured to train the machine learning from the information acquired by the sensor, and the escape route is calculated on the basis of the model.

10. The device of claim 8 further comprising:

a processor with a program to determine a presence of a threat based on information gathered by the sensor.

11. The device of claim 8 further comprising:

at least one of a visual display and an audible transducer to warn a person of a presence of a threat in the venue.

12. The device of claim 8 wherein:

the program is configured to train a model of the venue by machine learning from pre-acquired information accessed by a communication system.

13. The device of claim 12 wherein:

information accessed by the communication system is selected by use of the information acquired by the sensor.

14. The device of claim 8 wherein:

the output device comprises a vehicle to carry the person.

15. A method of escaping a threat in a venue comprising:

(a) gathering information about conditions in the venue with a sensor after a person enters the venue with the sensor;

(b) training a machine learning system implemented on a computer with the information,

(c) operating the machine learning system to produce a plan of escape from the venue subsequent to detection of a threat to safety of a person by the machine learning system;

(d) transmitting a step of the plan of escape with an output device; and

(e) performing the step of the plan of escape in accordance with the transmission.

16. The method of claim 15 wherein:

17. The method of claim 15 wherein:

the machine learning system is trained with at least one of data concerning venue layouts in a class of venues and routes for escape from venues in a class of venues.

18. The method of claim 15 wherein:

the machine learning system is trained prior to entry to the venue by a person to be protected with at least one of data concerning a layout of the venue, data concerning layouts of a group of venues selected to represent a class of venues, routes for escape from the venue and routes to escape from a group of venues selected to represent a class of venues, and

the machine learning system is trained with the information subsequent to entry of the person into the venue.

19. The method of claim 15 further comprising:

warning the person of the presence of the threat with the output device.

20. The method of claim 15 wherein:

the output device is a vehicle adapted to be directed autonomously by the step of the plan.