Abstract: Techniques are presented for the detection and management of collision warning (CW) events. A training dataset comprising videos of vehicle collisions and non-collisions, sensor readings, environmental conditions, and more is utilized to train a CW classification model for detecting potential collision events in vehicles. A backend CW classification model, with greater computational resources, employs a more complex neural network to review CW events received by the Behavioral Monitoring System (BMS) based on video data, achieving higher precision and reducing false positives. The CW model is installed in vehicles for real-time detection, while the backend model is deployed at the BMS. The BMS validates detected CW events, filters out false positives, and streamlines the review process for fleet administrators and customers. Additional BMS filtering operations include assessing non-proximity-related CW events and camera impairments, with the filtered CW events presented for review in the safety inbox.

Abstract: A method is disclosed that includes the step of: responsive to determining, based on input data, that a gateway system is receiving power from a power source of a mobile asset and the mobile asset is not in-use, transitioning the gateway system from a first mode to a second mode, wherein transitioning comprises turning on a functionality of the gateway system that is turned off when the gateway system is operating in the first mode.

Abstract: An improved system and method of defining speeding severity levels, receiving over a wireless network from a vehicle device speed data and metadata, determining associated speeding severity levels, based at least in part on the determined associated speeding severity levels, determining whether an alert communication is to be transmitted the vehicle device, and in response to determining that an alert communication is to be transmitted the vehicle device, transmitting the alert communication via a wireless interface to the vehicle device, the alert communication configured to cause the vehicle device to provide a corresponding in-vehicle alert.

Abstract: Techniques are presented for enabling vehicle immobilization based on location. One method includes providing a user interface to configure immobilization settings. The vehicle is configured for threat detection, and a data structure is created to define locations where immobilization is enabled. A Behavioral Monitoring System (BMS) prepares and sends a hierarchical map structure to the vehicle's communication device, which determines when to immobilize the vehicle based on its location. The communication device and engine immobilizer are installed in the vehicle. Continuous monitoring for tampering and jamming is performed, and upon detection, the method checks if geographic locking is enabled. If geographic locking is enabled, the vehicle is immobilized according to configuration rules, and the BMS is notified. The immobilization process can be ended based on predefined configuration parameters, such as a time interval, recovery of GPS signal, or driver notification.

Abstract: Methods, systems, and computer programs are presented for monitoring tailgating when a vehicle follows another vehicle at an unsafe distance. A method for enhancing a Following Distance (FD) machine learning (ML) model is disclosed. The method includes providing a management user interface (UI) for configuring FD parameters, followed by receiving FD events. A UI for manual FD annotation and another for customer review of filtered FD events are also provided. Annotations and customer review information are collected to improve the training set for the FD ML model. The FD model is then trained with the new data and downloaded to a vehicle. Once installed, the FD model is utilized to detect FD events within the vehicle, thereby enhancing the vehicle's safety and performance in driving scenarios by improving the accuracy and reliability of FD event predictions or detections.

Abstract: Methods, systems, and computer programs are presented for the assignment of courses based on driver behavior. One method includes monitoring driver behavior and analyzing reported events to generate course recommendations. These recommendations are based on configurable rules set by a fleet manager. A safety inbox user interface (UI) allows for the review of events and the assignment of one or more courses to a driver. The method further includes recording the assigned course as pending, notifying the driver, updating the driver application to reflect the course requirement, detecting course completion by the driver, and updating the Safety Behavior Management System (BMS) with the course completion record. This systematic approach ensures that drivers are aware of and complete necessary educational courses to improve driving safety and adherence to fleet standards.

Abstract: Disclosed are systems, methods, and non-transitory computer-readable media for customized vehicle operator workflows. A vehicle operator workflow includes a set of tasks to be completed by a vehicle operator. The vehicle operator workflow may be presented to the vehicle operator upon satisfaction of a triggering event, such as upon the vehicle operator logging into an account or the vehicle entering a specified geographic region. The vehicle operator workflow may enable the vehicle operator to complete the individual tasks as well as submit evidence that the tasks have been completed.

Abstract: A computer-implemented method for machine learning model operation can include, by one or more processors executing program instructions: providing a first training dataset comprising a plurality of images and associated object detection labels, providing a second training dataset comprising a plurality of images and associated classification labels, and providing a machine learning model comprising a model backbone, an object detection task head, and a classification task head. The method can further include training the machine learning model by training the object detection task head using the first training dataset and training the classification task head using the second training dataset. The method can further include deploying the trained machine learning model that includes the trained model backbone and the trained object detection task head but does not include the trained classification task head.

Abstract: Systems and methods are described for determining a peripheral's location based on observations from a plurality of centrals. A central may periodically transmit its location and observations of peripherals to a backend. The backend may match the location of the central with one or more observations of a peripheral to estimate a location of the peripheral. The backend may receive observations of the same peripheral from a plurality of centrals. The backend may aggregate estimated locations to triangulate a more precise location of the peripheral.

Abstract: Systems and methods are described for transmitting broadcasts by Peripherals, receiving the broadcasts by Centrals, and communicating with a Backend by Centrals. Specifically, a Peripheral may generate and transmit an encrypted broadcast packet to a Central, the Central may transmit the encrypted observation information to the Backend, which may decrypt the broadcast packet. Additionally, a Central may transmit a request for authorization to connect to a Peripheral to the Backend. If authorized by the Backend, the Central may connect to the Peripheral. Further, a Central may receive a plurality of broadcasts over a period of time. The Central may store information about the broadcasts and, at the conclusion of the period of time, transmit the information about the broadcasts and the last-received broadcast to the Backend.

Abstract: Methods, systems, and programs are presented for detecting impaired views in monitoring cameras. One method includes training a rotation classifier with unsupervised learning utilizing a first set of images. The rotation classifier is configured to receive an input image and generate a rotation feature embedding for the input image. In addition, the method includes training an impairment classifier with supervised learning utilizing a second set of images, impairment labels for each of the second set of images, and the rotation feature embedding, generated by the rotation classifier, for each of the second set of images. The method further includes accessing a vehicle image captured by a camera on a vehicle, and providing the vehicle image to the impairment classifier as input, and the impairment classifier outputs a camera impairment from a set of camera impairment categories. Further, the vehicle image and the camera impairment are presented on a user interface.

Abstract: An air quality notification system to perform operations that include: causing a sensor device to sample an air quality associated with a location at a predefined sampling interval; determining, at the sensor device, that a value corresponding with the air quality sampled by the sensor device transgresses a threshold value; causing the sensor device to transmit an alert to a server system responsive to the determining that the value corresponding with the air quality transgresses a threshold value; and causing the server system to present a notification at a client device responsive to the sensor device transmitting the alert, the notification including at least a display of an identifier associated with the location.

Abstract: A management server system may obtain sensor data generated by a plurality of sensors from a plurality of gateway devices. The sensor data may be associated with a plurality of physical assets. The management server system may generate a user interface for display by a user computing device based on the sensor data. The user interface may include visualizations of the plurality of physical assets and filters. The management server system can obtain a selection of a filter from the user computing device. Based on the selection of the filter, the management server system may filter the sensor data to identify a portion of sensor data. The management server system can filter the plurality of physical assets to identify a portion of the physical assets associated with the portion of sensor data. The management server system may generate a user interface including visualizations of the portion of physical assets.

Abstract: The technology disclosed herein enables configuration and sharing of a coverage map which displays asset data. In various implementations, an asset owner provides access to select portions of asset data, such as vehicle performance data, to an end-user. The asset owner configures a coverage map to display the selected data according to the type of asset data to be displayed, the scope of the data to be displayed, and other map attributes. Once configured, a URL is generated which the fleet operator provides to the end-user.

Abstract: Methods for performing operations for improving driver safety across a fleet of vehicles are disclosed. A plurality of safety events pertaining to a driving of a fleet of vehicles by a plurality of drivers are detected. A subset of the events is identified. The subset corresponds to one or more safety events of the plurality of safety events involving one or more vehicles of the fleet of vehicles to which drivers have not been assigned. A user interface is generated for presentation on a client device, the user interface including an interactive user interface element for accessing the subset of the events. One or more user interface elements are provided for accepting or rejecting recommendations for assignments of one of the plurality of drivers to each of the vehicles. The recommendations are generated based on an application of a machine-learned model to images of faces captured.

Abstract: A system receives vehicle metric data from a gateway device connected to a vehicle via a network connection. In a first power mode, the vehicle gateway device gathers vehicle metric data related to operation of the vehicle and/or location data and sends requested vehicle metric data to the system. Based on a lack of input, the vehicle gateway device enters a standby power mode to reduce the power consumption of the vehicle gateway device. The vehicle gateway device reduces the power consumption by reducing activity of the vehicle gateway device. Instead of reinitiating a network connection periodically to receive requests for vehicle metric data from the system, the vehicle gateway device maintains the network connection by sending keep alive messages between the vehicle gateway device and the system. The system uses the maintained network connection to request vehicle metric data.

Abstract: Disclosed are systems, methods, and non-transitory computer-readable media for managing diagnostics and malfunctions of a vehicle. A route management system detects an event representing an improper operation of an electronic logging device (ELD) of a vehicle; in response to detecting the event: generates, for display, a notification representing the event to a driver of the vehicle; retrieves instructions for resolving the improper operation of the ELD; presents the instructions as part of the notification; and coordinates communicating the event and resolution status of the event to a fleet manager of the vehicle.

Abstract: A driver monitoring system includes internal and external cameras attached to a vehicle. The cameras capture video of the vehicle's driver and the area surrounding the vehicle for later review by a coach or supervisor. To ensure privacy of people who appear in video, portions of the video may be processed to blur faces, license plates, and/or other features. Furthermore, access control mechanisms exist so that only users in specific roles have access to review certain types of video.

Abstract: In various implementations, an asset gateway service receives log messages comprising asset data captured by a gateway device associated with an asset. The log messages each include a first identifier corresponding to a first representation of the gateway device, which itself is associated with a first organization. The service persists data from the log to a first datastore accessible by the first organization. The service also determines whether a second representation of the gateway device (i.e., a clone) exists. If so, the service persists at least some of the data from the log message to a second datastore, which itself may be associated with a second organization, thereby allowing both organizations to access the data in a secure yet convenient manner.

Abstract: A system receives historical vehicle battery data from a gateway device connected to a vehicle. Some vehicles with plug-in rechargeable batteries recommend/require that the vehicle computer be turned off when recharging. Thus, obtaining a current state of charge while a vehicle is charging can be difficult because the vehicle computer can be off. While the vehicle/gateway device is unable to transmit current battery data, the systems estimate a battery charge from the historical data.