Systems and methods for autonomous navigation and mapping in various environments

Abstract

A novel approach is presented that enables human rescuers and human operators to explore and navigate treacherous and challenging environments safely and efficiently by providing an integrated, practical, flexible, and robust system. Leveraging advanced sensor deployment, adaptive communication networks, and real-time subject detection with AR-enabled navigation to enhance safety and real-time usability, a single practical solution is disclosed that is flexible, scalable, robust and error resilient, secure, energy-efficient, sustainable, and configured to optimize sensor placement and robot behavior to adapt to changes in the environment. Key innovations include federated learning for privacy-preserving model updates and multi-modal sensing for superior accuracy.

Description

SYSTEMS AND METHODS FOR AUTONOMOUS NAVIGATION AND MAPPING IN VARIOUS ENVIRONMENTSREEATED APPEICATIONS

[0001] This application claims the benefit of U.S. patent application 63 / 740,867 filed December 31 , 2024, which is incorporated by reference along with all other references cited in this application.BACKGROUND OF THE INVENTION

[0002] In scenarios such as natural disasters, military operations, or hazardous industrial environments, exploring unknown spaces can pose significant risks to human operators. Current technologies for autonomous navigation often face challenges in maintaining communication with the outside world, detecting the presence of a subject (e.g., detecting human presence), and providing reliable maps of the environment. Current systems suffer from limitations, including a reliance on pre-existing communication infrastructure, a limited ability to detect subjects in obstructed or cluttered environments, a lack of real-time adaptability in dynamic scenarios, and inadequate tools for human operator navigation assistance in hazardous spaces. Accordingly, there is a need for a robust system that integrates navigation, communication, mapping, and subject detection (e.g., human detection) into a single solution that reduces risks to human rescuers and human operators in hazardous or unknown environments and enables exploring such environments safely and efficiently.BRIEF DESCRIPTION OF THE DRAWINGS

[0003] Various embodiments of the invention are disclosed in the following detailed description and the accompanying drawings.

[0004] FIG. 1 shows a flowchart depicting an exemplary method for autonomous navigation and mapping in various environments.

[0005] FIG. 2 shows a system architecture flowchart according to some embodiments.

[0006] FIG. 3 shows a depiction of the operation of a human detection system with advanced sensing concepts according to some embodiments.

[0007] FIG. 4 shows a block diagram of an exemplary' computer system to perform portions of methods for autonomous navigation and mapping in various environments.

[0008] FIG. 5 shows a block diagram of an exemplary’ system for performing the method of FIG. 1 and other methods as described herein.DETAILED DESCRIPTION

[0009] The invention can be implemented in numerous ways, including as a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer readable storage medium; and / or a processor, such as a processor configured to execute instructions stored on and / or provided by a memory coupled to the processor. In this specification, these implementations, or any other form that the invention may take, may be referred to as techniques. In general, the order of the steps of disclosed processes may be altered within the scope of the invention. Unless stated otherwise, a component such as a processor or a memory described as being configured to perform a task may be implemented as a general component that is temporarily configured to perform the task at a given time or a specific component that is manufactured to perform the task.

[0010] A detailed description of one or more embodiments of the invention is provided below along with accompanying figures that illustrate the principles of the invention. The invention is described in connection with such embodiments, but the invention is not limited to any embodiment. The scope of the invention is limited only by the claims and the invention encompasses numerous alternatives, modifications and equivalents.Numerous specific details are set forth in the following description in order to provide a thorough understanding of the invention. These details are provided for the purpose of example and the invention may be practiced according to the claims without some or all of these specific details. For the purpose of clarity, technical material that is known in the technical fields related to the invention has not been described in detail so that the invention is not unnecessarily obscured.

[0011] As used herein, the term ‘‘module’' denotes a software component configured to implement a major function of a larger piece of software. For instance, a module may comprise a core component of a complex software system, a program that performs a core or essential function for other programs, a program that serves as the core foundation for a larger piece of software, a piece of software that is usually not found in isolation, or software that performs a specific or repetitive function. Classically, a module is packaged into a software library, though a module can also form part of an application programming interface (API). A module may alternatively be referred to as an “engine.”

[0012] As used herein, the term “processor” refers to one or more devices, circuits, and / or processing cores configured to process data, such as computer program instructions.

[0013] As used herein, the term “memory” refers to any device capable of storing electronic information that communicates with one or more processors to process data, suchas computer program instructions, modules, or engines.

[0014] As used herein, the term ’’or" refers to both its conjunctive and disjunctive meanings unless such a meaning is impossible. For instance, the phrase “A or B” shall generally be construed to mean A alone, B alone, and both A and B to the extent that any of these meanings is not impossible. As another example, the phrase “A, B, or C” shall generally be construed to mean A alone. B alone, C alone, A and B but not C, A and C but not B, B and C but not A, and all of A, B, and C to the extent that any of these meanings is not impossible.

[0015] In scenarios such as natural disasters, military operations, or hazardous industrial environments, exploring unknown spaces can pose significant risks to human operators. Current technologies for autonomous navigation often face challenges in maintaining communication with the outside world, detecting the presence of a subject (e.g., detecting human presence), and providing reliable maps of the environment. Current systems suffer from limitations, including a reliance on pre-existing communication infrastructure, a limited ability to detect subjects in obstructed or cluttered environments, a lack of real-time adaptability in dynamic scenarios, and inadequate tools for human operator navigation assistance in hazardous spaces. Accordingly, there is a need for a robust system that integrates navigation, communication, mapping, and subject detection (e.g., human detection) into a single solution that reduces risks to human rescuers and human operators in hazardous or unknown environments and enables exploring such environments safely and efficiently.

[0016] The disclosed techniques are directed to autonomous robotic systems, communication networks, and human detection technologies. Unlike current systems, the integrated solution described herein integrates navigation, communication, mapping, and subject detection (e.g., human detection) into a single solution that reduces risks to human rescuers and human operators in hazardous or unknown environments. In particular, the disclosed integrated solution combines advanced sensor deployment, adaptive communication networks, and real-time advanced subject detection, including human detection, with Augmented Reali ty (AR)-guided navigation that enables humans to explore challenging and treacherous environments safely, effectively, and efficiently using a single, robust and versatile system. By using non-human operators, machines, and / or robots to deploy sensors in various environments at least in some embodiments, this single, robust, and versatile system provides an integrated solution configured to reduce the risk of harm to human rescuers and human operators in hazardous or unknown environments and to enable exploring such environments safely and efficiently.

[0017] Systems and methods are disclosed for safely navigating and exploring unknown environments by using non-human operators, machines, and / or robots to deploy sensors that establish a communication network, build or generate a real-time map of the environment, and scan for the presence of a subject, including scanning for the presence of humans, with applications in search and rescue, military operations, and hazardous environment exploration. The disclosed integrated solution is configured to provide practical tools and guidance to human rescuers and human operators including by: enabling visualization of mapped pathways and subject locations in real time through the generation of a real-time map of the environment; identifying and marking human locations within a mapped space in order to facilitate immediate rescue or operational planning; and using laser projections to mark safe pathways for navigation in the environment. Other key innovations in the disclosed systems and methods include federated learning for privacy-preserving model updates and multi-modal sensing for superior accuracy.

[0018] In contrast to current and existing systems, the techniques disclosed herein enable human rescuers and human operators to explore and navigate treacherous and challenging environments safely and efficiently by providing an integrated, practical, flexible, and robust solution that can be used, for example, in search and rescue efforts to save and preserve lives.

[0019] A novel approach is presented that enables human rescuers and human operators to explore and navigate treacherous and challenging environments safely and efficiently by providing an integrated, practical, flexible, and robust system. Leveraging advanced sensor deployment, adaptive communication networks, and real-time subject detection with AR-enabled navigation to enhance safety and real-time usability, a single practical solution is disclosed that is flexible, scalable, robust and error resilient, secure, energy-efficient, sustainable, and configured to optimize sensor placement and robot behavior to adapt to changes in the environment. Key innovations also include federated learning for privacy-preserving model updates and multi-modal sensing for superior accuracy.

[0020] In particular, the disclosed integrated system for exploring unknown environments safely and efficiently imparts the following advantages and capabilities not currently provided in a single solution.

[0021] Enhanced Safety: Reduces the risk of harm to human rescuers and human operators in hazardous or unknown environments through the use of non-human operators, machines, and / or robots to deploy sensors in the unknown environments.

[0022] Real-Time Usability: Provides immediate data and guidance through ARtools and visual markers thus enabling human rescuers and human operators to navigate and move through unknown environments more efficiently, effectively, and safely.

[0023] Flexible and Scalable: Dynamically adapts to diverse environments with modular and upgradable components.

[0024] Robust and Error Resilient: Includes fail-safe mechanisms and adaptive algorithms to address mapping inaccuracies or unexpected obstacles in real-time and for communication and mapping redundancy, thereby increasing the reliability of the system which is especially critical in challenging environments for search and rescue missions. Identifies and responds to irregular signal patterns or environmental changes using anomaly detection models.

[0025] Secure: Ensures data integrity by applying differential privacy in federated learning models to protect sensitive information and using federated learning for privacypreserving model updates. Protects communication and data transfer using end-to-end encryption to ensure integrity and confidentiality of transmitted data (e.g., data transmitted between a robot and one or more sensors, and at least one external operator).

[0026] Energy Efficient: Uses energy-efficient protocols including low-power communication methods or sleep modes for sensors or sensor nodes in deployment and operation.

[0027] Sustainable: Provides enhanced sustainable operation using solar-powered or energy -harvesting components to reduce environmental impact and extend operational duration.

[0028] Optimized Using Machine Learning: Continuously improves system performance and accuracy using performance metrics and reinforcement learning algorithms or other optimization methods for closed-loop optimization of sensor placement and robot behavior, and by providing closed-loop feedback between observed network performance and deployment decisions. Optimizes network configurations and sensor placement by updating federated learning models over time without requiring centralized data storage.

[0029] FIG. 1 shows a flowchart depicting an exemplary method 100 for autonomous navigation and mapping in various environments. As shown in FIG. 1, a method 100 for autonomous navigation and mapping of an environment comprises deploying a robot into the environment at 110; deploying one or more sensors in strategic locations in the environment using the robot at 120; providing data transfer between the robot, the one or more sensors, and at least one external operator at 130; receiving data from the one or more sensors at 140: detecting a subject using at least some of the received data at 150: and building or generatinga map of the environment using at least some of the received data at 160. The subject in some cases can be human, while in other cases, the subject may be a non-human living organism or a non-living thing.

[0030] In some embodiments, the one or more sensors comprise one or more sensor nodes configured to establish a communication network. In some cases, at least one of the one or more sensor nodes is embedded with mesh network capabilities for communication and data relay. Mesh network capabilities can include, for example, one or more of: (i) the ability for a node to act as a router and / or repeater for other nodes; (ii) support for multi -hop routing between nodes that are not in direct radio range; (iii) automatic discovery' of neighboring nodes and topology; (iv) self-healing behavior, in which traffic is rerouted around failed or obstructed nodes; and (v) execution of one or more ad hoc routing protocols as known in the art (for example, AODV, OLSR, HWMP, RPL, or similar protocols). For a study of ad hoc networks and their different varieties, including wireless sensor networks, wireless mesh networks, see, e.g., R. Agrawal, et al., “Classification and comparison of ad hoc networks: A review.” Egyptian Informatics Journal, 24(1), 1-25. https: / / doi.Org / 10.1016 / j.eij.2022.10.004, which is incorporated by reference herein in its entirety'. For another survey of mesh networks, also see, e g., Y. Chai, and X. J. Zeng, “The development of green wireless mesh network: A survey. ” Journal of Smart Environments and Green Computing, vol. 1, no. 1, pp. 47-59, 2021 and I. F. Akyildiz and E.P. Stuntebeck, "Wireless underground sensor networks: Research challenges. ’’ Ad Hoc Networks, vol. 4, no. 6, pp. 669-686, 2006, each of which is incorporated by reference herein in its entirety.

[0031] In some embodiments, the one or more sensor nodes are configured to be deployable and attachable. In some embodiments, the one or more sensor nodes comprise one or more physically separate devices that are configured to remain within the environment after deployment. In some cases, the one or more sensor nodes are configured to be dropped, released, launched, or attached at various locations in an environment (e.g., floors, walls, ceilings, beams, pipes, and / or debris). The locations may be in an indoor space or may be outdoors, and may comprise man-made structures or natural environments. In certain implementations, the sensor nodes include mechanical attachment features or devices (e.g., magnets, adhesive pads, straps, hooks, barbs, or micro-spine structures) configured to secure the sensor nodes to irregular or overhead surfaces. Once deployed, the sensor nodes function as independent components of the communication network and continue to provide sensing data, including but not limited to RF measurements. Wi-Fi signal analysis, thermal readings, acoustic data, or environmental signals. In some embodiments, the sensor nodes are deployedat locations or in positions that are not physically accessible to human operators, including confined spaces, voids within debris, elevated structural elements, or unsafe regions of a hazardous environment. As described in more detail herein with respect to FIG. 2, in some embodiments one or more robots are configured to deploy or attach the one or more sensor nodes at various locations in an environment, thereby reducing risks to human rescuers and human operators in hazardous or unknown environments.

[0032] In some examples, and as described above, at least one of the one or more sensor nodes is configured to use WiFi sensing technologies, Wi-Fi signal analysis, radiofrequency (RF) sensing, and / or thermal imaging. For an overview of WiFi sensing technologies, Channel State Information (CSI) basics, activity recognition, human presence detection, and positioning, see, e.g. Y. Ma et al., "WiFi Sensing with Channel State Information: A Survey" ACM Comput. Surv., Vol. 1, No. 1, Article 1. Publication date: January 2019, which is incorporated by reference herein in its entirety.

[0033] In some applications, the subject is a human subject and the data is processed to detect a human presence. In some cases, multi-modal data is processed in real time using customized machine learning (ML) models and algorithms to improve detection accuracy and resource efficiency. Such machine learning models and algorithms include, for example, supervised learning (classification and regression), unsupervised learning (clustering and dimensionality reduction), semi-supervised learning, and reinforcement learning. Learningbased algorithms used for processing multi-modal data as described herein can include Decision Tree, Naive Bayes, Dynamic Time Wrapping, k Nearest Neighbor, Support Vector Machine, Self-Organizing Map, Hidden Markov Models, Convolutional / Recurrent Neural Network, and Long Short-Term Memory algorithms.

[0034] In some embodiments, ML models and algorithms are customized and configured to assign a priority level or triage classification to each detected subject based on one or more inferred characteristics derived from the received data. Such characteristics include, for example, motion patterns, estimated posture (e.g., upright, prone), thermal signatures, lack of detectable movement over time, proximity to identified hazards such as structural instability, high-temperature regions, and hazardous gases. In some embodiments, a real-time map may be augmented with visual indicators representing the determined triage classifications. Additionally, in some embodiments, an AR-Guided Rescue Workflow is provided through an AR system comprising AR navigation tools configured to display the triage indicators using distinct colors, shapes, icons, or labels to guide human operators toward higher-priority subjects. The AR system may further generate or project pathrecommendations that account for both the subject’s triage level and predicted hazard zones within the environment, thereby providing a unified decision-support workflow for rescue operations.

[0035] In some embodiments, advanced detection of the subject is performed using Received Signal Strength Indicator (RSSI), Channel State Information (CSI), Time of Flight (ToF), Frequency Modulated Carrier Wave (FMCW), Multipath Propagation, and / or Doppler Effect. As an example, RSSI is used to estimate distances between a router and a detected subject, CSI is used to track precise movements and locations of a subject within the environment, ToF is used to measure the time it takes for signals to travel and return from a detected subject; FMCW is used to map differences in time to the shifts of carrier frequency (and can be deployed to measure ToF of radio signals), Multipath Propagation is used to refine localization of a detected subject by analyzing signal bounces off walls and objects, and the Doppler Effect is used to detect motion of a subject based on frequency shifts. For a broad overview' and comparison of these and other wireless sensing technologies (including RSSI-based detection, CSI-based detection, FMCW, and Doppler Shift-based techniques) see, e.g. J. Liu et al., '"Wireless Sensing for Human Activity: A Survey f DOI10.1109 / COMST.2019.2934489, IEEE, Communications Surveys and Tutorials, which is incorporated by reference herein in its entirety7. For a description of device-free CSI presence detection and using non-linear techniques to improve CSI human presence detection accuracy, see, e.g.. S. Palipana et al., "Channel State Information Based Human Presence Detection!' 2016 ACM.ISBN978-1 -4503-4264-3 / 16 / 11, which is incorporated by reference herein in its entirety7.

[0036] Additionally, a sensor fusion algorithm is used in some embodiments to combine at least two of the modalities to improve accuracy and resilience in complex, cluttered, or obstructed spaces. In general, sensor fusion and multi-modal information fusion combine data from multiple sensors or multiple sources (e.g., images and LiDAR). In some embodiments, a processing module performs sensor fusion to combine data from multiple sensing modalities, including RF-based sensing (e.g., RSSI. CSI, ToF), lidar, cameras, thermal imagers, acoustic sensors, or inertial sensors. In some cases, Kalman filters, extended Kalman filters, unscented Kalman filters, particle filters, Bayesian filtering methods, or deep-leaming-based fusion networks are used to perform sensor fusion. Sensorfusion outputs may be used to refine subject localization, improve mapping quality, or enhance the robustness of the communication network.

[0037] A discussion of various techniques and an overview of some main data fusion approaches can be found in J. Queralta et al., “Collaborative Multi-Robot Search and Rescue: Planning, Coordination, Perception, and Active Piston." Digital Object Identifier10.1109 / ACCESS.2020.3030190, which is incorporated by reference herein in its entirety. Also see, e.g. Y. Ma et al., “WiFi Sensing with Channel State Information: A Survey,” cited above and incorporated by reference herein, for a summary and discussion of various algorithms that may be used for WiFi sensing and sensor fusion, including modeling-based algorithms (e.g., Theoretical Models: Fresnel Zone Model, Angle of Arrival / Departure, Time of Flight, Amplitude Attenuation, Phase Shift, Doppler Spread, Power Delay Profile, MultiPath Fading, Radio Propagation: Reflection. Refraction, Diffraction, Absorption, Polarization, Scattering; Statistical Models: Rician Fading, Power Spectral Density, Coherence Time / Frequency, Self / Cross Correlation; Algorithms: MUSIC, Thresholding, Peak / V alley Detection, Minimization / Maximization), learning-based algorithms (Decision Tree, Naive Bayes, Dynamic Time Wrapping, k Nearest Neighbor, Support Vector Machine, Self-Organizing Map, Hidden Markov Models, Convolutional / Recurrent Neural Network, and Long Short-Term Memory algorithms), and hybrid algorithms that combine the use of modeling-based and learning-based algorithms to improve detection, recognition, or estimation results.

[0038] For a discussion of wireless sensing, types of wireless signals, theoretical models, signal preprocessing techniques, activity segmentation, feature extraction, classification, and application, see, e.g., J. Liu et al., “Human Activity Sensing with Wireless Signals: A Survey,” Sensors (Basel). 2020 Feb 22;20(4):1210. doi: 10.3390 / s20041210. PMID: 32098392; PMCID: PMC7071003, which is incorporated by reference herein in its entirety. Also see, Y. Liu, et al.. “Harvesting Ambient RF for Presence Detection Through Deep Learning,” IEEE Trans Neural Netw Leam Syst. 2022 Apr;33(4): 1571-1583. doi: 10.1109 / TNNLS.2020.3042908. Epub 2022 Apr 4. PMID: 33361005, which is incorporated by reference herein in its entirety, for a discussion on the use of ambient radio frequency (RF) signals for human presence detection through deep learning, including for example Convolutional Neural Networks (CNN-based presence detection) using Wi-Fi and CSI.

[0039] In some examples, the communication network comprises an adaptive mesh network. In some cases, reinforcement learning algorithms are employed to optimize sensor node placement dynamically to maximize coverage of the environment and minimize signal interference.

[0040] In some embodiments, one or more performance metrics associated with communication network quality and subject detection performance are computed or determined. Such performance metrics may include, by way of example and not limitation: signal coverage, link quality, localization uncertainty, subject-detection confidence, or predicted hazard likelihood within unmapped regions of the environment. In some embodiments, based at least in part on the performance metrics, one or more candidate locations are identified for subsequent sensor node deployment or robot movement. In some embodiments, a reinforcement learning algorithm or other optimization method selects actions that improve one or more performance metrics over time, thereby establishing a closed-loop feedback system between observed network performance and deployment decisions. In some implementations, this optimization process continues dynamically as the robot navigates, sensor nodes are deployed, and additional sensor data is collected, enabling the system to adapt to evolving environmental conditions.

[0041] In some embodiments, as described above, reinforcement learning or other optimization methods are employed to select sensor-node placement locations and robot navigation actions thereby providing closed-loop optimization of sensor placement and robot behavior. Reinforcement-learning techniques used in the methods and systems as described herein include Q-leaming, Deep Q-Networks, policy-gradient methods, actor-critic methods, and multi-agent reinforcement learning. In some embodiments, a reward function is based on one or more performance metrics, including but not limited to signal coverage, link-quality metrics, localization uncertainty, subject-detection confidence, predicted hazard likelihood, or expected information gain derived from future sensing.

[0042] In some embodiments, predictive mapping algorithms are used to anticipate hazards and to improve the safety of operations. In some embodiments, predictive mapping is performed at least in part by using a simultaneous localization and mapping (SLAM) algorithm, which may include lidar-based SLAM, visual SLAM, RGB-D SLAM, inertial-aided SLAM, or hybrid algorithm approaches combining multi-modal data and / or multiple sensing modalities. In some embodiments, predictive mapping is performed using occupancygrid maps, voxel maps, or point-cloud representations. Exploration strategies include frontierbased exploration, information-theoretic exploration, or utility-based action selection to identify candidate navigation goals that expand the mapped region while maintaining communication with deployed sensor nodes. In some embodiments, and as described with respect to FIG. 2 herein, a real-time mapping engine is used to perform predictive mapping

[0043] In some embodiments, path-planning and hazard-aware navigation algorithms are used to determine or compute robot navigation paths or human-operator guidance paths using graph-based or sampling-based planning algorithms, such as A*, D*, D* Lite, RRT, RRT*, or their variants. Path costs may incorporate predicted hazards, structural instabilities, RF coverage quality, or subject-priority' levels determined by triage analysis as discussed above. The resulting paths may be provided to AR tools or laser-projection systems for realtime guidance of human operators.

[0044] For a discussion and overview of advances in deep learning including deep reinforcement learning (DRL) methods for autonomous robotic exploration and predictive mapping algorithms, see, e.g., J. Placed, et al., “A Survey on Active Simultaneous Localization and Mapping: State of the Art and New Frontiers" Survey Paper. March 12, 2023,https: / / natanaso.github.io / ref / Placed_ActiveSLAMSurv ey_TRO23.pdf?utm_source=chatgpt.c om, which is incorporated by reference herein in its entirety.

[0045] For a discussion of predictive mapping algorithms for simultaneous localization and mapping (SLAM) system-based indoor mapping using autonomous mobile robots in unknown environments, see, e.g., A. Eldemiry et al., 'Autonomous Exploration of Unknown Indoor Environments for High-Quality Mapping Using Feature-Based RGB-D SLAMf Sensors. 2022; 22(14):5117. https: / / doi.org / 10.3390 / s22145117, which is incorporated by reference herein in its entirety. Also see. e.g., Feng A. Xie Y, Sun Y, Wang X, Jiang B, Xiao J. Efficient Autonomous Exploration and Mapping in Unknown Environments. Sensors (Basel). 2023 May 15;23(10):4766. doi: 10.3390 / s23104766. PMID: 37430680; PMCID: PMC10221315, which is incorporated by reference herein in its entirety, for a further discussion of DRL and predictive mapping algorithms. In particular, the latter describes a Local-and-Global Strategy (LAGS) algorithm that combines a local exploration strategy with a global perception strategy.

[0046] In some embodiments, Augmented Reality (AR) tools configured for use by human operators are used to enable or facilitate visualization of mapped pathways and subject locations in real time. See, e.g. E. Argo et al., "Augmented Reality User Interfaces for First Responders: A Scoping Literature Review,” arXiv:2506.09236vl, which is incorporated by reference herein in its entirety, for a review of the current landscape of AR technologies designed for first responders. Additionally, a review of multi-robot systems supporting Search and Rescue (SAR) operations, with system-level considerations and focusing on the algorithmic perspectives for multi-robot coordination and perception is presented in J.Queralta et al., “Collaborative Multi-Robot Search and Rescue: Planning, Coordination, Perception, and Active Vision.” cited above and incorporated by reference herein in its entirety.

[0047] In some embodiments, tools configured for use by human operators can be used to provide laser projections or other means of marking as known in the art to mark safe pathways for navigation in the environment. Thus, in some cases, a generated map is provided to one or more tools configured for use by a human operator to assist one or more human operators to navigate through unknown or hazardous environments in real time. In some instances, the map is updated in real time, or according to a particular schedule, or based on or triggered by a key event.

[0048] Network configurations and sensor placement can be optimized using the techniques described herein by updating federated learning models over time without requiring centralized data storage. In particular, network configurations and sensor placement can be optimized using the techniques described herein by updating one or more federated learning models over time, without requiring raw data to be aggregated in a centralized repository. In such embodiments, individual devices, or sensor nodes, locally train model parameters using their own observations and periodically transmit model updates (e.g., gradients or weight deltas) to an aggregation service. The aggregated parameters are then redistributed to the nodes, thereby improving the global model while keeping local training data on-device. In addition, differential privacy mechanisms can be integrated into the federated learning process, for example by injecting calibrated random noise into model updates or applying other privacy -preserving transformations, so that the contribution of any single user, device, or location cannot be inferred from the shared parameters, thereby protecting sensitive information. Federated learning techniques are generally known in the art (see, e.g., McMahan et al., "Communication-Efficient Learning of Deep Networks from Decentrcdized Data,” Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS 2017), vol. 54 of Proceedings of Machine Learning Research, pp. 1273-1282, 2017; Kairouz et al., “Advances and Open Problems in Federated Learning,” Foundations and Trends in Machine Learning, vol. 14, nos. 1-2, pp. 1-210, 2021. doi: 10.1561 / 2200000083), each of which is incorporated by reference herein in its entirety. Moreover, differential privacy in the federated learning models can be applied to protect sensitive information. Differential privacy techniques are also well known in the art (see, e.g., Dwork et al. , “Calibrating Noise to Sensitivity in Private Data Analysis.” Theory of Cry ptography Conference (TCC 2006), Lecture Notes in Computer Science, vol. 3876, pp.265-284. Springer, Berlin, 2006. doi: 10.1007 / 11681878 14 ; Abadi et al., “ Deep Learning with Differential Privacy f Proceedings of the 2016 ACM SIGS AC Conference on Computer and Communications Security (CCS ’16), pp. 308-318. Association for Computing Machinery, New York, NY, USA, 2016. doi: 10.1145 / 2976749.2978318; Geyer et al., “Differentially Private Federated Learning f arXiv preprint arXiv:1712.07557, 2017), each of which is incorporated by reference herein in its entirety.

[0049] In some examples, fail-safe mechanisms and adaptive algorithms can be used at least in part to address mapping inaccuracies or unexpected obstacles in real-time.Furthermore, anomaly detection models are used in some embodiments to identify and respond to irregular signal patterns or environmental changes. Fail-safe mechanisms, adaptive algorithms, and anomaly detection models are well-known in the art. See, e.g., P. Nunes, et al., “Challenges in predictive maintenance - A review f CIRP Journal of Manufacturing Science and Technology, Volume 40, 2023, Pages 53-67 (discussing anomaly detection), which is incorporated by reference herein in its entirety7.

[0050] In certain cases, end-to-end encry ption is used to ensure integrity and confidentiality of data transmitted between the robot, the one or more sensors, and an external operator or operators. Energy-efficient protocols including, for example, low-power communication methods or sleep modes for sensors or sensor nodes may also be employed to improve or enhance system performance. End-to-end encry ption methods are well known in the art. See, e.g., E. Rescorla, E. “The Transport Layer Security (TLS) Protocol Version 1.3. ” RFC 8446, Internet Engineering Task Force (IETF), Aug. 2018 (defines modem end-to- end channel protection over TCP); and National Institute of Standards and Technology7(NIST), “Announcing the Advanced Encryption Standard (AES) ’’ FIPS PUB 197, Nov. 26, 2001 (defines AES, commonly used within TLS and other E2E schemes), each of which is incorporated by reference herein in its entirety7.

[0051] Solar-powered or other energy-harvesting components may also be employed to reduce environmental impact and extend operational duration. Energy-harvesting techniques are known in the art. See, e.g.. J. A. Paradiso and T. Stamer, “Energy scavenging for mobile and wireless electronics. "IEEE Pervasive Computing, vol. 4, no. 1, pp. 18-27, Jan.-Mar. 2005; S. Sudevalayam and P. Kulkami, “Energy harvesting sensor nodes: Survey and implications. ” IEEE Communications Survey s & Tutorials, vol. 13, no. 3, pp. 443-461, 2011; Priya, S., & Inman, D. J. (eds.). Energy Harvesting Technologies. New York, NY, USA: Springer, 2009, each of which is incorporated by reference herein in its entirety.

[0052] FIG. 2 shows a system architecture flowchart 200 according to some embodiments. As shown in FIG. 2, an Autonomous or Manual Robot 210 can be used to deploy one or more Deployable Sensor Nodes 220. In some embodiments. Autonomous or Manual Robot 210 is configured to deploy or attach the one or more Deployable Sensor Nodes 220 at various locations in an environment, thereby reducing risks to human rescuers and human operators in hazardous or unknown environments.

[0053] In some embodiments, the one or more Deployable Sensor Nodes 220 are configured to be deployable and attachable. In some embodiments, the one or more Deployable Sensor Nodes 220 comprise one or more physically separate devices that are configured to remain within the environment after deployment. In some cases, the one or more Deployable Sensor Nodes 220 are configured to be dropped, released, launched, or attached at various locations in an environment (e.g., floors, walls, ceilings, beams, pipes, and / or debris). The locations may be in an indoor space or may be outdoors, and may comprise man-made structures or natural environments.

[0054] In some embodiments, the Deploy able Sensor Nodes 220 include mechanical attachment features or devices (e.g., magnets, adhesive pads, straps, hooks, barbs, or microspine structures) configured to secure the sensor nodes to irregular or overhead surfaces. Once deployed, the Deployable Sensor Nodes 220 can function as independent components of the communication network and continue to provide sensing data, including but not limited to RF measurements, Wi-Fi signal analysis, thermal readings, acoustic data, or environmental signals. Examples of Deploy able Sensor Nodes 220 include Wi-Fi routers or custom-designed communication devices. In some embodiments, the Deployable Sensor Nodes 220 are deployed at locations or in positions that are not phy sically accessible to human operators, including confined spaces, voids within debris, elevated structural elements, or unsafe regions of a hazardous environment.

[0055] In some cases, Deployable Sensor Nodes 220 are used to establish a Mesh Communication Network 230 for robust communication and data relay. Additionally, Deployable Sensor Nodes 220 can be embedded with subject detection (e.g., human detection) technologies, such as signal interference analysis, RF sensing, and thermal imaging.

[0056] In some examples, and as described above, at least one of the one or more Deployable Sensor Nodes 220 is configured to use WiFi sensing technologies, Wi-Fi signal analysis, radio-frequency (RF) sensing, and / or thermal imaging. For an overview of WiFi sensing technologies, Channel State Information (CSI) basics, activity recognition, humanpresence detection, and positioning, see, e.g. Y. Ma et al., “WiFi Sensing with Channel State Information: A Survey" ACM Comput. Surv., Vol. 1, No. 1, Article 1. Publication date: January 2019, which is incorporated by reference herein in its entirety.

[0057] Autonomous or Manual Robot 210 can be fully autonomous or can be operated manually by a human operator. In addition, robot 210 can be equipped with mobility mechanisms suitable for diverse terrains (e.g., wheels, tracks, legs, propellers, wings, or any other propulsion device). Robot 210 can be mounted on or an integrated component of a mobile device or apparatus such as a vehicle configured to move on land, air, or water. Robot 210 can also be integrated with various tools, instruments, or sensors including but not limited to cameras, laser detection and ranging (LIDAR) sensors, ultrasonic sensors, and thermal imagers for navigation and mapping. Robot 210 is configured to and can deploy sensor nodes at strategic locations and is configured to and can provide data (including data from Sensor Nodes 220) to Real-Time Mapping Engine 240.

[0058] In some embodiments, Robot 210 is configured to and can deploy sensor nodes at strategic locations. Reinforcement learning or other optimization methods are used by the system (e.g., by the Processing Module as described with respect to FIG. 5) to select sensor-node placement locations and robot navigation actions thereby providing closed-loop optimization of sensor placement and robot behavior. Reinforcement-learning techniques used in the methods and systems as described herein include Q-leaming, Deep Q-Networks, policy-gradient methods, actor-critic methods, and multi-agent reinforcement learning. In some embodiments, a reward function is based on one or more performance metrics, including but not limited to signal coverage, link-quality metrics, localization uncertainty, subject-detection confidence, predicted hazard likelihood, or expected information gain derived from future sensing.

[0059] In some embodiments, Real-Time Mapping Engine 240 is configured to perform predictive mapping. In some embodiments, Real-Time Mapping Engine 240 is configured to use a simultaneous localization and mapping (SLAM) algorithm. In some embodiments, the SLAM algorithm comprises lidar-based SLAM, visual SLAM, RGB-D SLAM, inertial-aided SLAM, and / or hybrid algorithm approaches combining multi-modal data and / or multiple sensing modalities. In some embodiments, Real-Time Mapping Engine 240 is configured to perform predictive mapping using occupancy-grid maps, voxel maps, or point-cloud representations. Exploration strategies used by the Real-Time Mapping Engine 240 can include frontier-based exploration, information-theoretic exploration, or utility-basedaction selection to identify candidate navigation goals that expand the mapped region while maintaining communication with deployed sensor nodes.

[0060] In some embodiments, various sensing technologies (e.g., sensors or sensor nodes used or deployed by Robot 210 and / or Sensor Nodes 220) comprise motion sensors, vision-based sensors, acoustic-based sensors, and pyroelectric infrared (PIR) sensors. Such sensing systems can extract signal changes associated with human activities based on different sensing methods (e.g., RSSI, CSI, FMCW and Doppler Effect) see, e.g. J. Liu et al., “Wireless Sensing for Human Activity: A Survey. ” Additionally, received signals are available in most WiFi devices, and indicate the path loss of wireless signals with respect to a certain distance.

[0061] Additionally, a sensor fusion algorithm is used in some embodiments (for example, by the Processing Module described herein with respect to FIG. 5) to combine at least two of the modalities to improve accuracy and resilience in complex, cluttered, or obstructed spaces. In general, sensor fusion and multi-modal information fusion combine data from multiple sensors or multiple sources (e.g., images and LiDAR). In some embodiments, a processing module performs sensor fusion to combine data from multiple sensing modalities, including RF-based sensing (e.g., RSSI, CSI, ToF), lidar, cameras, thermal imagers, acoustic sensors, or inertial sensors. In some cases, Kalman filters, extended Kalman filters, unscented Kalman filters, particle filters, Bayesian filtering methods, or deep-leaming-based fusion networks are used to perform sensor fusion. Sensorfusion outputs may be used to refine subject localization, improve mapping qualify, or enhance the robustness of the communication network.

[0062] In some embodiments, Real-Time Mapping Engine 240 is part of a Processing Module as described in more detail herein with respect to FIG. 5. In some cases. Real-Time Mapping Engine 240 can be configured to use Advanced Human Detection Algorithms 250 in order to detect a human presence. The results of applying Advanced Human Detection Algorithms 250 can be used to inform tools (including for example, Augmented Reality or AR Tools 260) used by human operators. Such tools can be used to provide Laser-Guided Pathways 270 to be used by External Operators 280.

[0063] In some embodiments, Advanced Human Detection Algorithms 250 are performed using Received Signal Strength Indicator (RSSI), Channel State Information (CSI), Time of Flight (ToF), Frequency Modulated Carrier Wave (FMCW), Multipath Propagation, and / or Doppler Effect. As an example, RSSI is used to estimate distances between a router and a detected subject, CSI is used to track precise movements and locationsof a subj ect within the environment, ToF is used to measure the time it takes for signals to travel and return from a detected subject; FMCW is used to map differences in time to the shifts of carrier frequency (and can be deployed to measure ToF of radio signals), Multipath Propagation is used to refine localization of a detected subject by analyzing signal bounces off walls and objects, and the Doppler Effect is used to detect motion of a subject based on frequency shifts. For a broad overview and comparison of these and other wireless sensing technologies (including RSSI-based detection, CSI-based detection, FMCW, and Doppler Shift-based techniques) see, e.g. J. Liu et al., “Wireless Sensing for Human Activity: A Survey f DOI 10.1109 / COMST.2019.2934489, IEEE, Communications Surveys and Tutorials, which is incorporated by reference herein in its entirety. For a description of device-free CSI presence detection and using non-linear techniques to improve CSI human presence detection accuracy, see, e.g., S. Palipana et al., “Channel State Information Based Human Presence Detection^ 2016 ACM.ISBN978-1-4503-4264-3 / 16 / 11, which is incorporated by reference herein in its entirety.

[0064] In some applications, the subject is a human subject and the data is processed (for example by the Processing Module as described with respect to FIG. 5) to detect a human presence. In some cases, multi-modal data is processed in real time using customized machine learning (ML) models and algorithms to improve detection accuracy and resource efficiency. Such machine learning models and algorithms include, for example, supervised learning (classification and regression), unsupervised learning (clustering and dimensional ity reduction), semi-supervised learning, and reinforcement learning. Learning-based algorithms used for processing multi-modal data as described herein can include Decision Tree, Naive Bayes, Dynamic Time Wrapping, k Nearest Neighbor, Support Vector Machine, SelfOrganizing Map. Hidden Markov Models. Convolutional / Recurrent Neural Network, and Long Short-Term Memory algorithms.

[0065] In some embodiments, ML models and algorithms are customized and configured (for example, in the Processing Module as described with respect to FIG. 5) to assign a priority level or triage classification to each detected subject based on one or more inferred characteristics derived from the received data. Such characteristics include, for example, motion patterns, estimated posture (e.g., upright, prone), thermal signatures, lack of detectable movement over time, proximity to identified hazards such as structural instability, high-temperature regions, and hazardous gases. In some embodiments, a real-time map may be augmented with visual indicators representing the determined triage classifications.Additionally, in some embodiments, an AR-Guided Rescue Workflow is provided through anAR system comprising AR navigation tools configured to display the triage indicators using distinct colors, shapes, icons, or labels to guide human operators toward higher-priority subjects. The AR system may further generate or project path recommendations that account for both the subject’s triage level and predicted hazard zones within the environment, thereby providing a unified decision-support workflow for rescue operations.

[0066] FIG. 3 shows a depiction 300 of the operation of a human detection system with advanced sensing concepts according to some embodiments. As shown in FIG. 3, relative locations and placement of ROBOT 310, one or more SENSOR NODES (at 320, 321, 322 and 323 respectively) and two HUMAN TARGETS (at 330 and 331 respectively) are shown in ENVIRONMENT 301.

[0067] In this case. ROBOT 310 is depicted in FIG. 3 as deployed in an environment 301 on the display 300. Also shown in FIG. 3 are one or more sensor nodes 320, 321, 322, and 323 with circles around each sensor node to illustrate the coverage provided by each sensor node as deployed in ENVIRONMENT 301 relative to the location of ROBOT 310. The system and methods as disclosed herein provide coverage by the sensor nodes to obtain data that is used to scan for human presence and in this case, to detect HUMAN TARGET 330 and 331.

[0068] FIG. 4 is a block diagram of a computer system 400 used in some embodiments to perform portions of methods for autonomous navigation and mapping in various environments described herein (such as operation 110, 120. 130, 140, 150 or 160 of method 100 as described herein with respect to FIG. 1).

[0069] FIG. 4 illustrates one embodiment of a general-purpose computer system. Other computer system architectures and configurations can be used for carrying out the processing of the inventions described herein (including, for example, the Processing Module as described with respect to FIG. 5). In some embodiments, computer system 400 may be utilized as a component in systems for autonomous navigation and mapping in various environments as described herein. Computer system 400 is made up of various subsystems described below, includes at least one microprocessor subsystem 401. In some embodiments, the microprocessor subsystem comprises at least one central processing unit (CPU) or graphical processing unit (GPU). The microprocessor subsystem can be implemented by a single-chip processor or by multiple processors. In some embodiments, the microprocessor subsystem is a general-purpose digital processor which controls the operation of the computer system 400. Using instructions retrieved from memory 404. the microprocessor subsystem controls the reception and manipulation of input data, and the output and display of data onoutput devices.

[0070] The microprocessor subsystem 401 is coupled bi-directionally with memory 404, which can include a first primary storage, typically a random-access memory (RAM), and a second pri mary storage area, typically a read-only memory (ROM). As is well known in the art, primary storage can be used as a general storage area and as scratch-pad memory, and can also be used to store input data and processed data. It can also store programming instructions and data, in the form of data objects and text objects, in addition to other data and instructions for processes operating on microprocessor subsystem. Also, as well known in the art, primary storage typically includes basic operating instructions, program code, data and objects used by the microprocessor subsystem to perform its functions. Primary storage devices 404 may include any suitable computer-readable storage media, described below, depending on whether, for example, data access needs to be bi-directional or unidirectional. The microprocessor subsystem 401 can also directly and very rapidly retrieve and store frequently needed data in a cache memory (not show n).

[0071] A removable mass storage device 405 provides additional data storage capacity for the computer system 400, and is coupled either bi-directionally (read / write) or unidirectionally (read only) to microprocessor subsystem 401. Storage 405 may also include computer-readable media such as magnetic tape, flash memory, signals embodied on a carrier wave, PC-CARDS, portable mass storage devices, holographic storage devices, and other storage devices. A fixed mass storage 409 can also provide additional data storage capacity. The most common example of mass storage 409 is a hard disk drive. Mass storage 405 and 409 generally store additional programming instructions, data, and the like that typically are not in active use by the processing subsystem. It will be appreciated that the information retained within mass storage 405 and 409 may be incorporated, if needed, in standard fashion as part of primary storage 404 (e.g., RAM) as virtual memory.

[0072] In addition to providing processing subsystem 401 access to storage subsy stems, bus 406 can be used to provide access other subsystems and devices as well. In the described embodiment, these can include a display monitor 408, a network interface 407, a keyboard 402, and a pointing device 403, as well as an auxiliary input / output device interface, a sound card, speakers, and other subsystems as needed. The pointing device 403 may be a mouse, stylus, track ball, or tablet, and is useful for interacting with a graphical user interface.

[0073] The network interface 407 allows the processing subsystem 401 to be coupled to another computer, computer network, or telecommunications network using a networkconnection as shown. Through the network interface 407, it is contemplated that the processing subsystem 401 might receive information, e.g., data objects or program instructions, from another network, or might output information to another network in the course of performing the above-described method steps. Information, often represented as a sequence of instructions to be executed on a processing subsystem, may be received from and outputted to another network, for example, in the form of a computer data signal embodied in a carrier wave. An interface card or similar device and appropriate software implemented by processing subsystem 401 can be used to connect the computer system 400 to an external network and transfer data according to standard protocols. That is, method embodiments of the present invention may execute solely upon processing subsystem 401, or may be performed across a network such as the Internet, intranet networks, or local area networks, in conjunction with a remote processing subsystem that shares a portion of the processing. Additional mass storage devices (not shown) may also be connected to processing subsystem 401 through network interface 407.

[0074] An auxiliary I / O device interface (not shown) can be used in conjunction with computer system 400. The auxiliary I / O device interface can include general and customized interfaces that allow the processing subsystem 401 to send and, more typically, receive data from other devices such as microphones, touch-sensitive displays, transducer card readers, tape readers, voice or handwriting recognizers, biometrics readers, cameras, portable mass storage devices, and other computers.

[0075] In addition, embodiments of the present invention further relate to computer storage products with a computer readable medium that contains program code for performing various computer-implemented operations. The computer-readable medium is any data storage device that can store data which can thereafter be read by a computer system. The media and program code may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well known to those of ordinary skill in the computer software arts. Examples of computer-readable media include, but are not limited to, all the media mentioned above: magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CD-ROM disks; magneto-optical media such as floptical disks; and specially configured hardware devices such as application-specific integrated circuits (ASICs), programmable logic devices (PLDs), and ROM and RAM devices. The computer-readable medium can also be distributed as a data signal embodied in a carrier wave over a network of coupled computer systems so that the computer-readable code is stored and executed in a distributed fashion. Examples of program code include bothmachine code, as produced, for example, by a compiler, or fdes containing higher level code that may be executed using an interpreter. The computer system shown in FIG. 4 is but an example of a computer system suitable for use with the invention. Other computer systems suitable for use with the invention may include additional or fewer subsystems. In addition, bus 408 is illustrative of any interconnection scheme serving to link the subsystems. Other computer architectures having different configurations of subsystems may also be utilized.

[0076] FIG. 5 shows a block diagram of an exemplars’ System 500 for performing the method of FIG. 1. As shown in FIG. 5, in some embodiments, System 500 for autonomous navigation and mapping of an environment comprises: a Robot 510 configured to deploy one or more sensors in strategic locations in the environment; a Communication Module 530 configured to provide data transfer between the Robot 510, the one or more sensors, and at least one external operator (not shown); and a Processing Module 540 configured to: receive data provided by the one or more sensors; detect a subject (e.g., a human) using at least some of the received data; and build or generate a map of the environment using at least some of the received data. In some cases, the subject may be human, while in others, the subject may be a non-human living thing or non-living thing.

[0077] In some applications, the subject is a human subject and the Processing Module 540 is further configured to process received data and to detect a human presence. In some cases, Processing Module 540 is configured to process multi-modal data in real time using customized machine learning (ML) models and algorithms to improve detection accuracy and resource efficiency of the system 500. Such machine learning models and algorithms include, for example, supervised learning (classification and regression), unsupervised learning (clustering and dimensionality reduction), semi-supervised learning, and reinforcement learning. Learning-based algorithms used for processing multi-modal data as described herein can include Decision Tree, Naive Bayes, Dynamic Time Wrapping, k Nearest Neighbor, Support Vector Machine, Self-Organizing Map, Hidden Markov Models, Convolutional / Recurrent Neural Network, and Long Short-Term Memory' algorithms.

[0078] In some embodiments, the Processing Module 540 is configured to perform advanced detection of the subject (including human presence) using Received Signal Strength Indicator (RSSI), Channel State Information (CSI), Time of Flight (ToF), Frequency Modulated Carrier Wave (FMCW), Multipath Propagation, and / or Doppler Effect. As an example, RSSI is used to estimate distances between a router and a detected subject, CSI is used to track precise movements and locations of a subject within the environment, ToF is used to measure the time it takes for signals to travel and return from a detected subject;FMCW is used to map differences in time to the shifts of carrier frequency (and can be deployed to measure ToF of radio signals), Multipath Propagation is used to refine localization of a detected subject by analyzing signal bounces off walls and objects, and the Doppler Effect is used to detect motion of a subject based on frequency shifts. For a broad overview and comparison of these and other wireless sensing technologies (including RSSI-based detection, CSI-based detection, FMCW, and Doppler Shift-based techniques) see, e.g. J. Liu et al., ‘'Wireless Sensing for Human Activity: A Survey f DOI10.1109 / COMST.2019.2934489, IEEE, Communications Surveys and Tutorials, which is incorporated by reference herein in its entirety7. For a description of device-free CSI presence detection and using non-linear techniques to improve CSI human presence detection accuracy, see, e.g.. S. Palipana et al., "Channel State Information Based Human Presence Detection!' 2016 ACM.ISBN978-1-4503-4264-3 / 16 / 11, which is incorporated by reference herein in its entirety.

[0079] In some embodiments, the Processing Module 540 comprises Real-Time Mapping Engine 240 as described herein with respect to FIG. 2, which can be configured to use Advanced Human Detection Algorithms (shown at 250 in FIG. 2) in order to detect a human presence. The results of applying Advanced Human Detection Algorithms 250 can be used to inform tools (including for example, Augmented Reality7or AR Tools 260) used by human operators. Such tools, shown in FIG. 5 at 550. can be used to provide Laser-Guided Pathways 270 to be used by External Operators 280 as described with respect to FIG. 2. In some embodiments, the Real-Time Mapping Engine 240 functions as a control module to provide subject detection (e.g., human detection) and map generation functionalities. In some embodiments, the control module facilitates remote control or autonomous decision-making using Al algorithms.

[0080] In some embodiments. Processing Module 540 is configured to use a sensor fusion algorithm to combine at least two of the modalities as described herein to improve accuracy and resilience in complex, cluttered, or obstructed spaces or environments.

[0081] In some embodiments, Processing Module 540 is configured to perform sensor fusion by combining data from multiple sensing modalities, including RF-based sensing (e.g., RSSI, CSI, ToF), lidar, cameras, thermal imagers, acoustic sensors, or inertial sensors. In some cases, Kalman filters, extended Kalman filters, unscented Kalman filters, particle filters, Bayesian filtering methods, or deep-leaming-based fusion networks are used to perform sensor fusion. Sensor-fusion outputs may be used to refine subject localization, improve mapping quality', or enhance the robustness of the communication network.

[0082] In some embodiments, the communication network comprises an adaptive mesh network. In some cases. Processing Module 540 is configured to use reinforcement learning algorithms to optimize sensor node placement dynamically to maximize coverage of the environment and minimize signal interference. FIG. 3 provides a depiction of how the placement of sensor nodes in an environment determine and define the coverage of an environment with respect to detecting a human target or presence of a subject.

[0083] In some embodiments, the Processing Module 540 is configured to select sensor-node placement locations and robot navigation actions using reinforcement learning or other optimization methods to provide closed-loop optimization of sensor placement and robot behavior. Reinforcement-learning techniques include Q-leaming, Deep Q-Networks, policy-gradient methods, actor-critic methods, and multi-agent reinforcement learning. In some embodiments, a reward function is based on one or more performance metrics, including but not limited to signal coverage, link-quality metrics, localization uncertainty, subject-detection confidence, predicted hazard likelihood, or expected information gain derived from future sensing.

[0084] In some embodiments, the Processing Module 540 is configured to perform predictive mapping using predictive mapping algorithms to anticipate hazards and to improve the safety of operations. In some embodiments, predictive mapping is performed at least in part by using a simultaneous localization and mapping (SLAM) algorithm, which may include lidar-based SLAM, visual SLAM. RGB-D SLAM, inertial-aided SLAM, or hybrid algorithm approaches combining multi-modal data and / or multiple sensing modalities. In some embodiments, predictive mapping is performed using occupancy-grid maps, voxel maps, or point-cloud representations. Exploration strategies include frontier-based exploration, information-theoretic exploration, or utility-based action selection to identify candidate navigation goals that expand the mapped region while maintaining communication with deployed sensor nodes. In some embodiments, and as described with respect to FIG. 2 herein, a Real-Time Mapping Engine 240 is used to perform predictive mapping.

[0085] In some embodiments, the Processing Module 540 is configured to determine or compute robot navigation paths or human-operator guidance paths using path-planning and hazard-aware navigation algorithms. In some embodiments, the Processing Module 540 is configured to use graph-based or sampling-based planning algorithms, such as A*, D*, D* Lite, RRT, RRT*, or their variants. Path costs may incorporate predicted hazards, structural instabilities, RF coverage quality, or subject-priority levels determined by triage analysis as discussed above. In some embodiments, the Processing Module 540 is configured to providethe resulting paths may to AR tools or laser-projection systems to provide real-time guidance to human operators.

[0086] In some embodiments. Augmented Reality (AR) tools configured for use by human operators are used to enable or facilitate visualization of mapped pathways and subject locations in real time. See, e.g. E. Argo et al., "Augmented Reality User Interfaces for First Responders: A Scoping Literature Review.” arXiv:2506.09236vl, which is incorporated by reference herein in its entirety, for a review of the current landscape of AR technologies designed for first responders. Additionally, a review of multi-robot systems supporting Search and Rescue (SAR) operations, with system-level considerations and focusing on the algorithmic perspectives for multi-robot coordination and perception is presented in J.Queralta et al., “Collaborative Multi-Robot Search and Rescue: Planning, Coordination. Perception, and Active Vision,” cited above and incorporated by reference herein in its entirety.

[0087] In some embodiments, tools configured for use by human operators can be used to provide laser projections or other signals or means of marking as known in the art to mark safe pathways for navigation in the environment. Thus, in some cases, a generated map is provided to one or more tools configured for use by a human operator to assist one or more human operators to navigate through unknown or hazardous environments in real time. In some instances, the map is updated in real time, or according to a particular schedule, or based on or triggered by a key event.

[0088] In some embodiments, the Processing Module 540 is configured to assign a priority level or triage classification to each detected subject based on one or more inferred characteristics derived from the received data using customized ML models and algorithms. Such characteristics include, for example, motion patterns, estimated posture (e.g., upright, prone), thermal signatures, lack of detectable movement over time, proximity’ to identified hazards such as structural instability, high-temperature regions, and hazardous gases. In some embodiments, a real-time map may be augmented with visual indicators representing the determined triage classifications. Additionally, in some embodiments, an AR-Guided Rescue Workflow is provided through an AR system comprising AR navigation tools configured to display the triage indicators using distinct colors, shapes, icons, or labels to guide human operators toward higher-priority’ subjects. The AR system may further generate or project path recommendations that account for both the subject’s triage level and predicted hazard zones within the environment, thereby providing a unified decision-support workflow for rescue operations.

[0089] In some embodiments, the System 500 includes one or more Sensors 520 configured to be deployed by the Robot 510, and in some instances, the one or more Sensors 520 comprise one or more sensor nodes configured to establish a communication network. In some embodiments, the one or more Sensors 520 comprise the one or more Deployable Sensor Nodes 220 as described with respect to FIG. 2 and are configured to be deployable and attachable. In some embodiments, the one or more Sensors 520 comprise one or more physically separate devices that are configured to remain within the environment after deployment. In some cases, the one or more Sensors 520 are configured to be dropped, released, launched, or attached at various locations in an environment (e.g., floors, walls, ceilings, beams, pipes, and / or debris). The locations may be in an indoor space or may be outdoors, and may comprise man-made structures or natural environments.

[0090] In some embodiments, the Sensors 520 include mechanical attachment features or devices (e.g., magnets, adhesive pads, straps, hooks, barbs, or micro-spine structures) configured to secure the sensor nodes to irregular or overhead surfaces. Once deployed, the Sensors 520 can function as independent components of the communication network and continue to provide sensing data, including but not limited to RF measurements, Wi-Fi signal analysis, thermal readings, acoustic data, or environmental signals. Examples of Sensors 520 include Wi-Fi routers or custom-designed communication devices. In some embodiments, the Sensors 520 are deployed at locations or in positions that are not physically accessible to human operators, including confined spaces, voids within debris, elevated structural elements, or unsafe regions of a hazardous environment.

[0091] In some embodiments, at least one of the Sensors 520 is embedded with mesh network capabilities for communication and data relay. In some cases, at least one of the Sensors 520 is configured to use Wi-Fi signal analysis, RF sensing, and / or thermal imaging. In some cases. Sensors 520 are used to establish a mesh communication network (e.g.. Mesh Communication Network 230) for robust communication and data relay. Additionally, Sensors 550 can be embedded with subject detection (e.g., human detection) technologies, such as signal interference analysis, RF sensing, and thermal imaging.

[0092] In some cases, Robot 510 can be fully autonomous or can be operated manually by a human operator. In addition, Robot 510 can be equipped with mobility mechanisms suitable for diverse terrains (e.g., wheels, tracks, legs, propellers, wings, or any other propulsion device). Robot 510 can be mounted on or an integrated component of a mobile device or apparatus such as a vehicle configured to move on land. air. or water.Robot 510 can also be integrated with various tools, instruments, or sensors including but notlimited to cameras. LIDAR sensors, ultrasonic sensors, and thermal imagers for navigation and mapping. Robot 510 is configured to and can deploy one or more Sensors 520 (including for example, sensor nodes at strategic locations) and is configured to and can provide data (including data from sensor nodes) to Real-Time Mapping Engine 240.

[0093] Examples of sensor nodes include Wi-Fi routers or custom-designed communication devices. In some cases, sensor nodes are used to establish a communication network (e.g., a Mesh Communication Network 230 for robust communication and data relay as described with respect to FIG. 2). Additionally, sensor nodes can be embedded with subject detection (e.g., human detection) technologies, such as signal interference analysis, RF sensing, and thermal imaging.

[0094] In some examples, the environment can be unknown and the map of the environment can be a real-time 3D map. Additionally, Processing Module 540 can be configured to identify and mark human locations within a mapped space to facilitate immediate rescue or operational planning. Processing Module 540 can also be configured in some instances to use predictive mapping algorithms to anticipate hazards and to improve the safety of operations.

[0095] In some embodiments, Tools 550 such as Augmented Reality (AR) Navigation tools configured for use by human operators can be used to enable or facilitate visualization of mapped pathways and subject locations in real time. As an example, AR glasses or headmounted displays provide real-time visualization of the environment, including laser projections, mapped pathways, and detected subject (e g., human) locations. In particular, laser projection systems can mark safe pathways within hazardous environments.

[0096] In some cases, a generated map is provided (e.g., by the Real-Time Mapping Engine 240 or control module) to one or more tools configured for use by a human operator to assist one or more human operators to navigate through unknown or hazardous environments in real time. In some instances, the map is updated in real time, or according to a particular schedule, or based on or triggered by a key event.

[0097] In some embodiments, Processing Module 540 is configured to optimize network configurations and sensor placement by updating federated learning models over time without requiring centralized data storage. Processing Module 540 can also be configured to apply differential privacy in the federated learning models to protect sensitive information.

[0098] In some embodiments, Processing Module 540 is configured with fail-safe mechanisms and adaptive algorithms at least in part to address mapping inaccuracies orunexpected obstacles in real-time. In some cases, Processing Module 540 is configured to use anomaly detection models to identify and respond to irregular signal patterns or environmental changes.

[0099] In some embodiments, Communication Module 530 is configured to protect communication and data transfer using end-to-end encry ption to ensure integrity and confidentiality of data transmitted between the Robot 510, the one or more Sensors 520, and at least one external operator (not shown).

[0100] In some embodiments, System 500 is configured to use energy-efficient protocols in its deployment and operation. The protocols in some cases comprise low-power communication methods or sleep modes for sensors or sensor nodes. In some examples, solar-powered or energy -harvesting components are employed to reduce environmental impact and extend operational duration.

[0101] As described with respect to FIGS. 1-5, the disclosed systems and methods provide an integrated solution that combines autonomous navigation, real-time mapping, subject detection (e.g., human detection), and dynamic communication networks. In particular, unlike existing systems that focus on isolated functionalities, the disclosed systems and methods introduce a number of advantageous integrated features, including but not limited to the following:Dynamic Sensor Deployment and Adaptive Communication Network:

[0102] Sensor nodes are deployed dynamically by the robot or robotic system to create communication network (e.g., an adaptive mesh network). This ensures continuous connectivity, even in signal-blocked or complex environments, without relying on preexisting infrastructure. Reinforcement learning algorithms are employed to optimize node placement dynamically, maximizing coverage and minimizing signal interference.Dual-Function Sensor Nodes:

[0103] Sensor nodes serve dual purposes: maintaining robust communication and scanning for the presence of a subject (e.g., human presence) using technologies like Wi-Fi signal analysis, RF sensing, and thermal imaging. These nodes employ custom ML models to process multi-modal data in real time, improving detection accuracy and resource efficiency.Advanced Subject Detection Capabilities:

[0104] The system leverages advanced concepts and modalities including:i. Received Signal Strength Indicator (RSSI): To estimate distances between the router and a subject.ii. Time of Flight (ToF): To measure the time it takes for signals to travel and return.iii. Channel State Information (CSI): To track precise movements and locations within an environment.iv. Multipath Propagation: To refine subject localization by analyzing signal bounces off walls and objects.v. Doppler Effect: To detect motion of a subject based on frequency shifts.For search and rescue missions, the subject is often a human but the system can also be used to detect and locate non-human subjects or inanimate objects.Autonomous Navigation with Real-Time Mapping:

[0105] The system autonomously navigates through unknown environments while generating a real-time 3D map. Simultaneously, it identifies and marks subject locations (e.g., human locations) within the mapped space, facilitating immediate rescue or operational planning. Predictive mapping algorithms anticipate hazards, such as unstable structures or spreading fires, ensuring safer operations. Such maps can be provided or displayed in tools or other devices configured for use by human operators.

[0106] Reinforcement learning algorithms are employed to optimize node placement dynamically, maximizing coverage and minimizing signal interference.

[0107] A sensor fusion algorithm combines advanced concepts and modalities (e.g. Received Signal Strength Indicator (RSSI), Channel State Information (CSI), Time of Flight (ToF), Frequency Modulated Carrier Wave (FMCW), Multipath Propagation, and / or Doppler Effect) for superior accuracy and resilience in cluttered or obstructed spaces. In some cases, RSSI is used to estimate distances between a router and a detected subject, CSI is used to track precise movements and locations of a subject within the environment, ToF is used to measure the time it takes for signals to travel and return from a detected subject; FMCW is used to map differences in time to the shifts of carrier frequency (and can be deployed to measure ToF of radio signals). Multipath Propagation is used to refine localization of a detected subject by analyzing signal bounces off walls and objects, and the Doppler Effect is used to detect motion of a subject based on frequency shifts.AR and Laser-Guided Subject Navigation:

[0108] Augmented Reality (AR) tools, such as glasses or head-mounted displays, enable rescuers to visualize mapped pathways and subject locations (e.g.. human locations) in real time using data provided by sensors, which is received and processed by the systemsdescribed herein. For example, the processing module (comprising the Real-Time Mapping Engine or control module) can receive and process sensor data to generate a real-time 3D map that can be provided to a human operator through a tool. Additionally, laser projections can be used to mark safe pathways for navigation, bridging the gap between robotic exploration and human operator usability. These tools are supported by explainable Al systems, ensuring transparency in decision-making.Scalability and Adaptability:

[0109] The system dynamically adapts to different environments, from urban areas to subterranean spaces, by optimizing network configurations and sensor placement. Federated learning updates models over time without requiring centralized data storage, ensuring privacy and scalability.Error Handling and Redundancy:

[0110] The system includes fail-safe mechanisms, such as rerouting communication through alternative nodes if one fails, and adaptive algorithms to address mapping inaccuracies or unexpected obstacles in real-time. Anomaly detection models identify and respond to irregular signal patterns or environmental changes.Security and Data Privacy:

[0111] Communication is protected using end-to-end encry ption, ensuring the integrity and confidentiality of data transmitted between nodes, the robot, and external operators. Differential privacy is applied in federated learning models to protect sensitive information.Energy Efficiency and Sustainability:

[0112] The system employs energy-efficient protocols, such as low-power communication methods and sleep modes for sensor nodes. Solar-powered or energyharvesting components reduce environmental impact and extend operational duration.Embodiment Examples:Example Case 1

[0113] As an example, a system for autonomous navigation and mapping in various environments may operate and be deployed as follows.Autonomous Navigation

[0114] A robot is introduced into an unknown environment and begins exploring the environment autonomously or under manual control. Sensor nodes are strategically deployed by the robot using reinforcement learning algorithms to create an optimal communicationnetwork.Mapping

[0115] The robot collects spatial data using LIDAR sensors, cameras, and other onboard sensors. A real-time 3D map is generated by the processing module (e.g., using a real-time mapping module or a control module) and the map is updated as the robot navigates, with predictive algorithms identifying potential hazards.Subject Detection

[0116] Sensor nodes and the robot scan the environment for the presence of a subject (e.g., human presence) using a combination of:■ RSSI to gauge proximity.■ ToF for precise distance measurements.■ CSI for detailed motion tracking.■ Multipath Propagation for accurate localization in cluttered spaces.■ Doppler Effect to identify movement patterns

[0117] Detected subjects (e.g.. detected humans) are marked on the map by the realtime mapping or control module, with ML models classifying their activity (e.g., moving, unconscious) to prioritize rescue efforts.Communication and Navigation:

[0118] Video and audio feeds are relayed to external operators through the sensor network. AR tools provide navigational assistance to human operators entering the environment.Example Case 2

[0119] In a collapsed building, the robot autonomously deploys sensor nodes as it navigates. These nodes form a mesh network, ensuring uninterrupted communication. Using Wi-Fi signal interference, thermal imaging, and detection algorithms, the system detects trapped survivors, marking their locations on a 3D map. Human operators and rescuers equipped with AR glasses see real-time visual markers of safe paths and survivor locations. Laser projections guide the human operators through the debris, significantly reducing response time and risk. Predictive mapping highlights areas of potential collapse, ensuring rescuer safety.

Claims

1. CLAIMS1. A system for autonomous navigation and mapping of an environment comprising:a robot configured to deploy one or more sensors in strategic locations in the environment;a communication module configured to provide data transfer between the robot, the one or more sensors, and at least one external operator; anda processing module configured to:receive data provided by the one or more sensors;detect a subject using at least some of the received data; and generate a map of the environment using at least some of the received data.

2. The system of claim 1, further comprising one or more sensors configured to be deployed by the robot.

3. The system of claim 2, wherein the one or more sensors comprise one or more sensor nodes configured to establish a communication netw ork,4. The system of claim 3, wherein at least one of the one or more sensor nodes is embedded with mesh network capabilities for communication and data relay.

5. The system of any one of claims 3-4, wherein at least one of the one or more sensor nodes is configured to use at least one technology7from the group consisting of: Wi-Fi signal analysis, radio-frequency (RF) sensing, and thermal imaging.

6. The system of any one of claims 3-5, wherein the processing module is configured to process received data and detect a human presence.

7. The system of any one of claims 3-6, wherein the received data comprises multimodal data and wherein the processing module is configured to improve detection accuracy and resource efficiency by processing the multi-modal data in real time using custom machine learning (ML) models.

8. The system of any one of claims 3-7, wherein the processing module is configured to perform advanced detection of the subject using at least one modality' from the group consisting of: Received Signal Strength Indicator (RSSI), Channel State Information (CSI), Time of Flight (ToF), Frequency Modulated Carrier Wave (FMCW), Multipath Propagation, and Doppler Effect.

9. The system of claim 8, wherein RSSI is used to estimate distances between a router and a detected subject, wherein ToF is used to measure the time it takes for signals totravel and return from a detected subject; wherein CSI is used to track precise movements and locations of a subject within the environment; wherein multipath propagation is used to refine localization of a detected subject by analyzing signal bounces off walls and objects; and wherein the Doppler Effect is used to detect motion of a subject based on frequency shifts.

10. The system of claim 8, wherein the processing module is configured to use a sensor fusion algorithm to combine at least two of the modalities to improve accuracy and resilience in complex, cluttered, or obstructed spaces.

11. The system of any one of claims 1-10, wherein the subject comprises a human.

12. The system of any one of claims 3-11, wherein the communication network comprises an adaptive mesh network.

13. The system of any one of claims 1-12, wherein the processing module is configured to use reinforcement learning algorithms to optimize sensor node placement dynamically in order to maximize coverage of the environment and minimize signal interference.

14. The system of any one of claims 1-12, wherein the environment is unknown and wherein the map of the environment comprises a real-time three-dimensional map.

15. The system of claim 13, wherein the processing module is configured to identify and mark human locations within a mapped space in order to facilitate immediate rescue or operational planning.

16. The system of any one of claims 1-12. wherein the processing module is configured to use predictive mapping algorithms to anticipate hazards and to improve the safety of operations.

17. The system of any one of claims 1-16, further comprising Augmented Reality (AR) tools configured for use by human operators to display the generated map to enable visualization of mapped pathways and subject locations in real time.

18. The system of any one of claims 1-17, further comprising tools configured to receive data from the processing engine for use by human operators to provide laser projections to mark safe pathways for navigation in the environment.

19. The system of claim 3, wherein the processing module is configured to optimize the communication network and the strategic locations of sensor placement by updating federated learning models over time without requiring centralized data storage.

20. The system of claim 19, wherein the processing module is configured to apply differential privacy in the federated learning models to protect sensitive information.

21. The system of any one of claims 1-20, wherein the processing module is configured with fail-safe mechanisms and adaptive algorithms at least in part to address mapping inaccuracies or unexpected obstacles in real-time.

22. The system of any one of claims 1-21, wherein the processing module is configured to use anomaly detection models to identify and respond to irregular signal patterns or environmental changes.

23. The system of any one of claims 1-22, wherein the communication module is configured to protect communication and data transfer using end-to-end encryption to ensure integrity and confidentiality of data transmitted between the robot, the one or more sensors, and the at least one external operator.

24. The system of any one of claims 1-23. wherein the system is configured to use energy -efficient protocols in its deployment and operation.

25. The system of claim 24, wherein the protocols comprise low-power communication methods or sleep modes for sensors or sensor nodes.

26. The system of claims 24 or 25. further comprising using solar-powered or energyharvesting components to reduce environmental impact and extend operational duration.

27. A method for autonomous navigation and mapping of an environment comprising:deploying a robot into the environment;deploying one or more sensors in strategic locations in the environment using the robot;providing data transfer between the robot, the one or more sensors, and at least one external operator;receiving data from the one or more sensors;detecting a subject using at least some of the received data; and generating a map of the environment using at least some of the received data.

28. The method of claim 27, wherein the one or more sensors comprise one or more sensor nodes configured to establish a communication network,29. The method of claim 28, wherein at least one of the one or more sensor nodes is embedded with mesh network capabilities for communication and data relay.

30. The method of any one of claims 28-29, wherein at least one of the one or more sensor nodes is configured to use at least one technology from the group consisting of: Wi-Fi signal analysis, radio-frequency (RF) sensing, and thermal imaging.

31. The method of any one of claims 27-30, further comprising processing at least some of the received data to detect a human presence.

32. The method of claim 31, wherein the received data comprises multi-modal data and further comprising processing multi-modal data in real time using custom machine learning (ML) models.

33. The method of any one of claims 27-32, further comprising performing advanced detection of the subject using at least one modality from the group consisting of: Received Signal Strength Indicator (RSSI), Channel State Information (CSI), Time of Flight (ToF), Frequency Modulated Carrier Wave (FMCW), Multipath Propagation, and Doppler Effect.

34. The method of claim 33, wherein RSSI is used to estimate distances between a router and a detected subject; CSI is used to track precise movements and locations of a subject within the environment; ToF is used to measure the time it takes for signals to travel and return from a detected subject; FMCW is used to map differences in time to the shifts of carrier frequency’; Multipath Propagation is used to refine localization of a detected subject by analyzing signal bounces off walls and objects; and the Doppler Effect is used to detect motion of a subject based on frequency shifts.

35. The method of claim 33, further comprising using a sensor fusion algorithm to combine at least two of the modalities to improve accuracy and resilience in complex, cluttered, or obstructed spaces.

36. The method of any one of claims 27-35, wherein the subject comprises a human.

37. The method of any one of claims 28-36, wherein the communication network comprises an adaptive mesh network.

38. The method of any one of claims 28-37, further comprising using reinforcement learning algorithms to optimize sensor node placement dynamically in order to maximize coverage of the environment and minimize signal interference.

39. The method of any one of claims 28-38, wherein the environment is unknown and wherein the map of the environment comprises a real-time three-dimensional map.

40. The method of claim 36, further comprising identifying and marking human locations within a mapped space in order to facilitate immediate rescue or operational planning.

41. The method of any one of claims 27-40, further comprising using predictive mapping algorithms to anticipate hazards and to improve the safety of operations.

42. The method of any one of claims 27-41, further comprising using Augmented Reality (AR) tools configured for use by human operators to display the generated map to enable visualization of mapped pathways and subject locations in real time.

43. The method of any one of claims 27-42, further comprising using tools configured to receive data from the processing engine for use by human operators to provide laser projections to mark safe pathways for navigation in the environment.

44. The method of claim 28, further comprising optimizing the communication network and the strategic locations of sensor placement by updating federated learning models over time without requiring centralized data storage.

45. The method of claim 44, further comprising applying differential privacy in the federated learning models to protect sensitive information.

46. The method of any one of claims 27-45, further comprising using fail-safe mechanisms and adaptive algorithms at least in part to address mapping inaccuracies or unexpected obstacles in real-time.

47. The method of any one of claims 27-46, further comprising using anomaly detection models to identify and respond to irregular signal patterns or environmental changes.

48. The method of any one of claims 27-47, further comprising using end-to-end encry ption to ensure integrity' and confidentiality' of data transmitted between the robot, the one or more sensors, and the at least one external operator.

49. The method of any one of claims 27-47. further comprising using energy-efficient protocols.

50. The method of claim 49, wherein the protocols comprise low-power communication methods or sleep modes for sensors or sensor nodes.

51. The method of any one of claims 27-50, further comprising using solar-powered or energy -harvesting components to reduce environmental impact and extend operational duration.

52. A non-transitory computer readable medium storing thereon a computer program for autonomous navigation and mapping of an environment, which when executed by a computer, performs a method comprising:deploying a robot into the environment;deploying one or more sensors in strategic locations in the environment using the robot;providing data transfer between the robot, the one or more sensors, and at least one external operator;receiving data from the one or more sensors;detecting a subject using at least some of the received data; and generating a map of the environment using at least some of the received data.

Images