Generation of cellular world model and trajectory planning based thereon

Abstract

The present disclosure is related to devices and methods for generation of a cellular world model and trajectory planning based thereon. A method for generating a cellular world model includes: generating a cellular world model indicating a belief about what a true cellular environment is. A method for planning a trajectory for a communication device includes: planning a trajectory for the communication device in a cellular environment based on at least a cellular world model that indicates a belief about what the cellular environment is.

Description

[0001]GENERATION OF CELLULAR WORLD MODEL ANDTRAJECTORY PLANNING BASED THEREON Technical Field The present disclosure is related to the field of telecommunication, and inparticular, to devices and methods for generation of a cellular world model andtrajectory planning based thereon. Background The world is witnessing a widespread and increasing use of drones, or more technically the Unmanned (or Uncrewed) Aerial Vehicles (UAVs), in many segments of the economy and in our daily life. There are numerous use cases of UAVs in industry, goods transportation and delivery, surveillance, media production, etc. Fig. 1 is a diagram illustrating an exemplary scenario where a drone 100, whichalso functions as a User Equipment (UE), is flying across areas served by multipleGround Base Stations (GBSs), or more generally, Base Stations (BSs) 105-1 through 105-3. In Fig. 1, vertical white sticks show macro cell antennas attached to a GBS, andellipses show signal strength in that direction - between ellipses are antenna nulls withvery weak signal. As shown in Fig. 1, the drone UE 100 passes through many strong andweak signal paths from multiple antennas as it flies. As will be used hereinafter, thestrength of the signal from an antenna to a UAV at a point in space will be referred to asthe link gain to that point in space. In a setting where a UAV or drone UE is flying between 50m-300m altitude, it istasked to fly from a start position to a goal position, or to perform a more complexmission description. This is illustrated in Fig. 1. Typically, these UAVs find it difficult toreliably maintain a connection to cellular networks, particularly to Long Term Evolution(LTE) or suburban / rural 5th Generation (5G) networks. This is because of two physicalphenomena: UAVs likely have line-of-sight to many macro cell towers, meaning they canreceive much more interference than a ground user. The Signal-to-Interference-plus-Noise Ratio (SINR) needs to be above a certain minimum level (often taken as -6 dB) fora high enough connection quality to support a usable drone control link. The strength ofthe signal path from UAV to GBS cell antenna is difficult to predict, and can vary widelyin a very short distance: many deployed macro cell antennas have nulls in their link gainspatial distribution (see Fig. 1). If a UAV flies into a null, its connection strength to thatantenna drops very quickly.Typically, a UAV is served by a single cell (or a corresponding antenna) at any onetime. When the signal strength from another cell (or a corresponding antenna) issufficiently stronger than the current associated antenna (e.g., by a handover offset dB) for a sufficient length of time (e.g., measured by a handover timer), a handoverprocedure occurs to move the UAV to be served by the new, stronger cell. For groundusers travelling slowly (<100 km / h), handovers happen infrequently. A successfulhandover typically interrupts a cellular user’s connection for ~50 milliseconds. A failedhandover typically interrupts a cellular user’s connection for ~100 to 1000 milliseconds.It can happen if the link quality drops too quickly before the handover timer is finished. Both successful and unsuccessful handovers happen much more frequently todrone UEs, due to the physical phenomena described above. Numbers from publishedliteratures suggest that even if all handovers are successful, up to 0.5% of the time thecellular link will be interrupted by handover. Simulations carried out in a study suggestthat failed handovers happen frequently (~60% of all handovers at 150m altitude).Assuming ~2 seconds for network reconnection time after a failed handover, thesenumbers result in the cellular link being unavailable ~10% of the time. Summary Fig. 2A is a diagram illustrating an exemplary straight-line trajectory from a startposition to a goal position, while Fig. 2B is a diagram illustrating which cell the UAV isassociated to at each point in time, with time in seconds on the x axis. Please note thatFig. 2A only shows the cell having the strongest signal strength for each position, andtherefore a same cell may have multiple non-connected parts shown in Fig. 2A. Forexample, the cell 205-2 may have two non-connected parts shown in Fig. 2A, forexample, because its signal strength is lower than those of other cells (e.g., the cell 205-3, the cell 205-6) at the positions shown as belonging to the cells 205-3 and 205-6. As shown in Fig. 2A, the UAV is flying from the start position to the goal position by passing through the cells 205-1, 205-2, 205-3, 205-2 again, and finally the cell 205-4. This is also reflected in Fig. 2B. Referring to Fig. 2B, for ~15% of this trajectory, the UAV is not associated with a cell, for example, because it has failed a handover. Further,between ~35 to ~40 seconds, it is unable to connect to the network at all, during whichnone of the cells has a high enough SINR (e.g., -6 dB) or link (path) gain as also shown in Fig. 2B. If one had perfect knowledge of the antenna’s link gain spatial distribution at the UAV’s altitude, one could design a UAV motion plan that stays within high SINR areas, minimizes the number of handovers required, and avoids failed handovers with high probability. There are existing methods for such trajectory planning, trying to solve the aboveissues, and some of them are listed and discussed below and incorporated herein byreference in their entireties: [1] “Cellular-enabled UAV communication: A connectivity-constrained trajectoryoptimization perspective”, by Shuowen Zhang, Yong Zeng, and Rui Zhang, IEEETransactions on Communications 67, no. 3 (2018), Pages 2580-2604; [2] “Interference management for cellular-connected UAVs: A deep reinforcement learning approach”, by Ursula Challita, Walid Saad, and Christian Bettstetter, IEEE Transactions on Wireless Communications 18, no. 4 (2019), Pages 2125-2140; and [3] “Cellular-connected UAV trajectory design with connectivity constraint: A deepreinforcement learning approach”, by Yunfei Gao, Lin Xiao, Fahui Wu, Dingcheng Yang,and Zhongxiang Sun, IEEE Transactions on Green Communications and Networking 5, no. 3 (2021), pages 1369-1380. Several existing methods (e.g., Reference [1] and Reference [2]) implicitlyassume they have “perfect knowledge” of the cellular environment, by using a cellular simulator as a stand-in for the real world. They then use this cellular simulator to planUAV paths through high SINR areas. Further, a small number of methods (e.g.,Reference [3]) are able to collect data during each UAV mission, and use this data to learn to take better trajectories in the future. However, all existing methods either ignore handover entirely (by assuming thatthe UAV always instantaneously associates with the antenna having the highest link gainat the UAV’s current position), or present an oversimplified assumption of how handoverworks. For example, Reference [1] assumes that handover happens instantaneously at afixed radius from a GBS. Further, it also assumes that there is no handover betweenantennas of the same GBS, and therefore it does not model handover between antennason the same GBS. For another example, Reference [2] assumes that handover isinstantaneous, and the handover decision is made by the UAV rather than the cellularnetwork. Further, no existing method considers handover-induced communicationinterruption as part of their method or evaluation. More widely, many existing methods aim to optimize aspects of the cellular network such as antenna tilt angles, in order to reduce the number of handovers and ensure handovers are successful. One method optimizes antenna tilt angles to improve UAV handover, but it assumes that tilt angles can be changed unrealistically frequently (on a second-to-second basis rather than hours to days), and does not consider the impact on ground users. No existing methods optimize UAV trajectories to achieve better handover performance. It may also be possible to jointly optimize tilt angles and UAV trajectories. This isa harder problem because it requires evaluating the antenna tilt impact on ground users as well as on the UAV. Because tilt angles can only be changed infrequently, UAVtrajectory optimization is expected to be a better method of improving handoverperformance. Problems with methods that ignore / simplify handover Because existing methods ignore the communication interruption impacts of handover and make large simplifying assumptions on when / where handover happens,trajectories they generate will perform poorly if carried out in the real world: themethods make no attempt to minimize the number of handovers; the methods do notdifferentiate between successful and failed handovers, so do not attempt to ensure handovers are successful. All known existing methods evaluate their methods in simulation, where theirsimulators use the same assumptions as their methods do. This means that their resultsare overly optimistic: in a real-world deployment they will likely result in trajectories with an excessive number of handovers and a high proportion of failed handovers. Problems with methods that assume access to a perfect simulation of the real cellular network Even if their cellular simulators properly modeled handover behavior, there is noway to ensure that the cellular model is a perfect representation of the real network thatthe UAV will be deployed in. It is possible to construct accurate site-specific cellularmodels using detailed measurements and simulations. Building such a model requiresdetailed knowledge. Required information includes: antenna locations, antenna down-tilt angles, and antenna model numbers. It may also require assumptions and / or models of building geometry / locations, building materials and geographic area topography. Site- specific models may also require real measurements from the environment to evaluateor tweak the model. Therefore, it is impractical to build a site-specific model for alllocations that a UAV might operate in. The learning-from-data methods (e.g. Reference[3]) are deep reinforcement learning methods which require carrying out a large number of similar trajectories (thousands of trajectories with the same start / goal positions for Reference [3]) before they start achieving reasonable performance. Alternative / future approaches: Vertical-beamforming 5G massive Multiple-Input-Multiple-Output (mMIMO) effectively alleviates connection issues caused by antenna nulls, but these antennas are unlikely to be deployed outside relatively dense urban areas for economic reasons, forexample, as discussed in a white paper at https: / / www.ericsson.com / en / reports-and-papers / white-papers / advanced-antenna-systems-for-5g-networks, which is alsoincorporated herein by reference in its entirety. Many deployed networks still usestandard macro cell antennas, which have large antenna nulls. Therefore, to address or at least partially alleviate one or more of the aboveissues, some embodiments of the present disclosure are provided. According to a first aspect of the present disclosure, a method for generating acellular world model is provided. The method comprises: generating a cellular worldmodel indicating a belief about what a true cellular environment is. Further, some otherembodiments of the first aspect will be provided in the Detailed Description below. According to a second aspect of the present disclosure, a device for generating acellular world model is provided. The device comprises: a processor; a memory storinginstructions which, when executed by the processor, cause the device to: generate acellular world model indicating a belief about what a true cellular environment is. Insome embodiments, the instructions, when executed by the processor, further cause thedevice to perform any of the methods of the first aspect.According to a third aspect of the present disclosure, a method for planning atrajectory for a communication device is provided. The method comprises: planning atrajectory for the communication device in a cellular environment based on at least acellular world model that indicates a belief about what the cellular environment is.Further, some other embodiments of the third aspect will be provided in the Detailed Description below. According to a fourth aspect of the present disclosure, a device for planning atrajectory for a communication device is provided. The device comprises: a processor; amemory storing instructions which, when executed by the processor, cause the device to:plan a trajectory for the communication device in a cellular environment based on at least a cellular world model that indicates a belief about what the cellular environment is.In some embodiments, the instructions, when executed by the processor, further causethe device to perform any of the methods of the third aspect.According to a fifth aspect of the present disclosure, a computer programcomprising instructions is provided. The instructions, when executed by at least oneprocessor, cause the at least one processor to carry out any of the methods of any ofthe first aspect and the third aspect. According to a sixth aspect of the present disclosure, a carrier containing thecomputer program of the fifth aspect. In some embodiments, the carrier is one of anelectronic signal, optical signal, radio signal, or computer readable storage medium. According to a seventh aspect of the present disclosure, a system for trajectoryplanning is provided. The system comprises: one or more communication devices; and a trajectory planning device for planning a trajectory for at least one of the communicationdevices. In some embodiments, the trajectory planning device comprises: a processor; amemory storing instructions which, when executed by the processor, cause thetrajectory planning device to: plan a trajectory for the at least one communication devicein a cellular environment based on at least a cellular world model that indicates a beliefabout what the cellular environment is. In some embodiments, the system furthercomprises: a model generating device for generating the cellular world model, the devicecomprising: a processor; a memory storing instructions which, when executed by theprocessor, cause the model generating device to: generate the cellular world model. Insome embodiments, the instructions stored on memory of the model generating device,when executed by the processor of the model generating device, further cause themodel generating device to perform any of the methods of the first aspect. In someembodiments, the instructions stored on the memory of the trajectory planning device,when executed by the processor of the trajectory planning device, further cause thetrajectory planning device to perform any of the methods of the third aspect.Brief Description of the Drawings Fig. 1 is a diagram illustrating an exemplary scenario where a drone UE is flyingacross areas served by multiple BSs. Fig. 2A and Fig. 2B are diagrams illustrating an exemplary straight-line trajectoryfrom a start position to a goal position and corresponding cell association / SINR / link gainfor a UAV flying along the trajectory. Fig. 3A and Fig. 3B are diagrams illustrating an exemplary trajectory that is planned according to an embodiment of the present disclosure and corresponding cellassociation / SINR / link gain for a UAV flying along the trajectory.Fig. 4 is a diagram illustrating an exemplary system for trajectory planning according to an embodiment of the present disclosure. Fig. 5 is a diagram illustrating an exemplary procedure for generating a cellularworld model according to an embodiment of the present disclosure.Fig. 6 is a diagram illustrating an exemplary procedure for planning a trajectoryfor a UAV based on a cellular world model according to an embodiment of the presentdisclosure. Fig. 7 is a diagram illustrating an exemplary method for creating linearinterpolators for multiple per-cell link gain models according to an embodiment of thepresent disclosure. Fig. 8 is a diagram illustrating an exemplary cellular transition system according toan embodiment of the present disclosure.Fig. 9 is a diagram illustrating an exemplary Monte Carlo Tree Search (MCTS) search tree for trajectory planning according to an embodiment of the present disclosure. Fig. 10 is a diagram illustrating an exemplary simulation of a motion actionbetween states according to an embodiment of the present disclosure. Fig. 11 is a diagram illustrating an exemplary rollout of a trajectory to a goal position from a state according to an embodiment of the present disclosure. Fig. 12 is a flow chart illustrating an exemplary method for generating a cellularworld model according to an embodiment of the present disclosure.Fig. 13 is a flow chart illustrating an exemplary method for planning a trajectory for a communication device according to an embodiment of the present disclosure. Fig. 14 schematically shows an embodiment of an arrangement which may beused in devices according to an embodiment of the present disclosure.Fig. 15 shows an exemplary communication system in accordance with someembodiments. Fig. 16 shows an exemplary UE in accordance with some embodiments.Fig. 17 shows an exemplary network node in accordance with some embodiments.Fig. 18 is a block diagram illustrating an exemplary virtualization environment inwhich functions implemented by some embodiments may be virtualized. Detailed Description Hereinafter, the present disclosure is described with reference to embodiments shown in the attached drawings. However, it is to be understood that those descriptions are just provided for illustrative purpose, rather than limiting the present disclosure. Further, in the following, descriptions of known structures and techniques are omitted soas not to unnecessarily obscure the concept of the present disclosure.Of course, the present disclosure may be carried out in other specific ways than those set forth herein without departing from the scope and essential characteristics of the disclosure. One or more of the specific processes discussed below may be carried out in any electronic device comprising one or more appropriately configured processing circuits, which may in some embodiments be embodied in one or more application- specific integrated circuits (ASICs). In some embodiments, these processing circuits may comprise one or more microprocessors, microcontrollers, and / or digital signal processors programmed with appropriate software and / or firmware to carry out one or more of the operations described above, or variants thereof. In some embodiments, these processing circuits may comprise customized hardware to carry out one or more of the functions described above. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive. Although multiple embodiments of the present disclosure will be illustrated in the accompanying Drawings and described in the following Detailed Description, it should be understood that the disclosure is not limited to the disclosed embodiments, but instead is also capable of numerous rearrangements, modifications, and substitutions without departing from the present disclosure that as will be set forth and defined within the claims. Further, please note that although the following description of some embodimentsof the present disclosure is given in the context of 5G System (5GS), the presentdisclosure is not limited thereto. In fact, as long as generation of cellular world modeland / or trajectory planning are involved, the inventive concept of the present disclosuremay be applicable to any appropriate communication architecture, for example, to Global System for Mobile Communications (GSM) / General Packet Radio Service (GPRS), Enhanced Data Rates for GSM Evolution (EDGE), Code Division Multiple Access (CDMA),Wideband CDMA (WCDMA), Time Division - Synchronous CDMA (TD-SCDMA),CDMA2000, Worldwide Interoperability for Microwave Access (WiMAX), Wireless Fidelity(Wi-Fi), 4th Generation Long Term Evolution (LTE), LTE-Advanced (LTE-A), or 5G-Advanced (5GA), 6th generation (6G) mobile system standard, etc. Therefore, one skilledin the arts could readily understand that the terms used herein may also refer to their equivalents in any other infrastructure. For example, the term “device” used herein may refer to a server, a node, a UE, a terminal device, a mobile device, a mobile terminal, a mobile station, a user device, a user terminal, a wireless device, a wireless terminal, atransmission reception point (TRP), a base station, a base transceiver station, an accesspoint, a hot spot, a NodeB, an Evolved NodeB (eNB), a gNB, a network function, anetwork node, a network element, a satellite, an aircraft, or any other equivalents. Further, in some embodiments, when antennas are discussed hereinafter, theyare typically referring to a macro cell antenna, generally located at a height of 20-40m and serving relatively wide areas compared to micro-cells or smaller cells. In someembodiments, a drone UE is most likely to be served by these antennas, as smaller cellssuch as micro / pico-cells have smaller power budgets and smaller antennas. Further, although some embodiments are described with reference to drones / UAVs, the present disclosure is not limited thereto. In some other embodiments, the solutions described with reference to drones / UAVs may also be applicable to other movable entities, such as, a mobile phone, a portable digital assistant (PDA), a tablet, a laptop, a vehicle-mounted device, a bike, a vehicle, an aircraft, a ship, a submarine, aballoon, or any other entity for which a trajectory can be planned.Further, although the concept of cellular world model is described in the context of trajectory planning in some embodiments, the present disclosure is not limited thereto. In some other embodiments, the cellular world model may be used for other purposethan trajectory planning. For example, it may be used for network optimization (e.g.,how base stations / antennas are deployed to provide good coverage at a low cost) or anyother scenario where modeling a cellular environment is needed.Fig. 4 is a diagram illustrating an exemplary system 40 for trajectory planningaccording to an embodiment of the present disclosure. As shown in Fig. 4, the system 40may comprise one or more UEs 100-1 and 100-2 (collectively, UE(s) 100), a RadioAccess Network (RAN) 110 comprising one or more RAN nodes 105, a Core Network (CN)120, a trajectory planning device 140, a Cellular World Model (CWM) generating device145, and an Unmanned Traffic Management (UTM) server 150, which are communicatively coupled to each other via one or more links and / or networks, e.g., the Internet 130. However, the present disclosure is not limited thereto. In some other embodiments, the one or more entities in the system 40 may not be present, and / or additional entities may be comprised in the system 40. In other words, the system 40 may comprise additional entities, less entities, or some variants of the existing entitiesshown in Fig. 4. For example, when no traffic management is needed, the system 40may comprise no UTM 150 or any other traffic management system. For another example, when a constantly updated cellular world model is not required, the system 40may comprise no CWM generating device 145. For yet another example, the trajectoryplanning device 140 may communicate with the UE 100 directly without the Internet 130.Further, although two UEs 100 and one RAN node 105 are shown in Fig. 4, the presentdisclosure is not limited thereto. In some other embodiments, any number of UEs and / or any number of RAN nodes may be comprised in the system 40. In some embodiments, the one or more UEs 100 may comprise a UAV 100-1 anda mobile phone 100-2. However, the present disclosure is not limited thereto. In someother embodiments, the system 40 may comprise any type of UEs, for example, avehicle UE, an Internet of Things (IoT) UE. In some embodiments, the RAN node 105could be a base station (BS), a Node B, an evolved NodeB (eNB), a gNB, or an AN nodewhich provides the UEs 100 with access to the network. In some embodiments, the CWM generating device 145 and the trajectoryplanning device 140 may be collocated, or may be located separately from each other.For example, when the latest CWM is required (e.g., in a fast changing cellularenvironment), the CWM generating device 145 and the trajectory planning device 140may be collocated such that the CWM can be updated and put in use in time. Forexample, when the latest CWM is not required (e.g., in a slowly changing cellularenvironment), the CWM generating device 145 and the trajectory planning device 140may be located separately, or the CWM generating device 145 may be omittedcompletely, such that less power, network and processing resources can be consumedby not maintaining the latest CWM. Further, in some embodiments, the CWM generatingdevice 145 and / or the trajectory planning device 140 may be collocated with the UE 100. As shown in Fig. 4, the UEs 100 may be communicatively connected to the RANnode 105 which in turn may be communicatively connected to the CN 120 and then theInternet 130, such that the UEs 100 may finally communicate with other devices insideand / or outside the system 40, for example, the trajectory planning device 140, the CWMgenerating device 145, and / or the UTM 150. As mentioned above, all existing methods for trajectory planning whilemaintaining a connection to cellular networks either ignore handover entirely or presentan oversimplified assumption of how handover works. That is, no existing methodconsiders handover-induced communication interruption as part of their methods orevaluations. More widely, many existing methods aim to optimize aspects of the cellularnetwork such as antenna tilt angles, in order to reduce the number of handovers and ensure handovers are successful. One method optimizes antenna tilt angles to improve UAV handover, but it assumes that tilt angles can be changed unrealistically frequently, and does not consider the impact on ground users. No existing methods optimize UAVtrajectories to achieve better handover performance. It may also be possible to jointlyoptimize tilt angles and UAV trajectories. This is a harder problem because it requires evaluating the antenna tilt impact on ground users as well as on the UAV. Because tilt angles can only be changed infrequently, UAV trajectory optimization is expected to be a better method of improving handover performance. Because existing methods ignore the communication interruption impacts of handover and make large simplifying assumptions on when / where handover happens, trajectories they generate will perform poorly if carried out in the real world: Problems with methods that ignore / simplify handover Because existing methods ignore the communication interruption impacts ofhandover and make large simplifying assumptions on when / where handover happens,trajectories they generate will perform poorly if carried out in the real world: themethods make no attempt to minimize the number of handovers; the methods do not differentiate between successful and failed handovers, so do not attempt to ensure handovers are successful. All known existing methods evaluate their methods in simulation, where theirsimulators use the same assumptions as their methods do. This means that their resultsare overly optimistic: in a real-world deployment they will likely result in trajectories with an excessive number of handovers and a high proportion of failed handovers. Problems with methods that assume access to a perfect simulation of the real cellular network Even if their cellular simulators properly modeled handover behavior, there is no way to ensure that the cellular model is a perfect representation of the real network thatthe UAV will be deployed in. It is possible to construct accurate site-specific cellularmodels using detailed measurements and simulations. Building such a model requires detailed knowledge. Required information includes: antenna locations, antenna down-tilt angles, and antenna model numbers. It may also require assumptions and / or models of building geometry / locations, building materials and geographic area topography. Site- specific models may also require real measurements from the environment to evaluateor tweak the model. Therefore, it is impractical to build a site-specific model for alllocations that a UAV might operate in. The learning-from-data methods (e.g. Reference [3]) are deep reinforcement learning methods which require carrying out a large number of similar trajectories(thousands of trajectories with the same start / goal positions for Reference [3]) beforethey start achieving reasonable performance. Alternative / future approaches: Vertical-beamforming 5G massive Multiple-Input-Multiple-Output (mMIMO) effectively alleviates connection issues caused by antenna nulls, but these antennas are unlikely to be deployed outside relatively dense urban areas for economic reasons, forexample, as discussed in a white paper at https: / / www.ericsson.com / en / reports-and-papers / white-papers / advanced-antenna-systems-for-5g-networks, which is alsoincorporated herein by reference in its entirety. Many deployed networks still usestandard macro cell antennas, which have large antenna nulls. Therefore, to address or at least partially alleviate one or more of the aboveissues, some embodiments of the present disclosure are provided. In some embodiments, an objective is to plan a trajectory that minimizes a UAV’sdisconnection time ^ from the cellular network during a given mission. This trajectorycould be fixed before the flight (pre-planned), or the UAV could change its trajectory online using measurements collected during the flight (reactive). UAV is typicallydisconnected while its SINR is lower than SINRmin, which is a user-set parameter, orduring handover interruption time as described above. SINRmin is a user-settableparameter defining the quality of service required. The minimum required communications bit rate may define the minimum required SINR. In some embodiments, the disconnection time requirement could be given indifferent formats. For example: minimize total disconnection time divided by totalmission time; ensure with probability p that the disconnection time never exceedsmaximum disconnection time T; minimize some trajectory cost function based on a combination of trajectory total time-to-goal and disconnection time.In some embodiments, a method for trajectory planning is proposed as follows.Rather than assuming access to a perfectly accurate cellular simulator, data may be usedto build a spatial model of each antenna / cell’s link gain to a UAV at any given point inspace. In some embodiments, the per-antenna / per-cell model may be built using RSRPmeasurements recorded during UAV flights. In some embodiments, these RSRP measurements may be produced at ~1 Hz bythe cellular modem, for the purposes of making handover decisions. By recording theseRSRP measurements and the locations the measurements were taken, a spatial model oflink gain may be built for each antenna / cell. In some embodiments, the spatial link gain distribution may be modelled to eachantenna / cell with a Gaussian process (GP). In some embodiments, a GP kernel that isinspired by common antenna radiation patterns is designed. The GP kernel design resultsin a better GP model fitting compared to common GP kernels.In some embodiments, this collection of models and variable values are called asthe “cellular world model”. In some embodiments, the cellular world model mayrepresent the belief about what the true cellular environment (the “cellular world”,described in more detail below) looks like, given the dataset. In some embodiments,given any trajectory in space, the cellular world model can be used to predict both connection quality and handover behaviour along that trajectory. In some embodiments, a trajectory generator may use these predictions togenerate cellular-robust UAV trajectories. In some embodiments, to generate trajectories, an existing algorithm which isdesigned for decision-making without full knowledge or visibility of the environment maybe used. This means that the UAV’s trajectory should be robust to the true cellular behaviour it encounters during its mission. In some embodiments, readily available measurements that the cellular UAV isalready taking may be used to build a single cellular world model which can predict atleast two things: connection quality and handover dynamics. Therefore, someembodiments of the present disclosure provide a UAV trajectory optimization methodthat is robust to failed handovers and limits the number of handovers required toachieve its mission, and this UAV trajectory optimization method is the first method of itskind. Building models of RSRP / link gain, including building separate models for eachemitter / antenna / cell to predict both A) link gain and B) SINR for UAV trajectoryoptimization is a novel method. A) and B) are the two dynamics that contribute most topoor cellular UAV performance. However, the present disclosure is not limited thereto. Asmentioned above, in some other embodiments, other dynamics than A) and / or B) maybe predicted by the models. In some embodiments, an automated system is proposed, which may recordreadily-available RSRP measurements the UAV is already taking, and use these tobuild / update a cellular world model, which may be later used for planning future UAVmissions. In some embodiments, the way a probabilistic model of the link gain spatialdistribution from an antenna / cell is generated is proposed.With some embodiments of the present disclosure, the following advantages can be achieved. 1. The information needed to generate UAV trajectories can be learned from datawithout having to build a complex cellular model of the area the UAV is flying in. Noneed to know GBS locations, antenna model numbers, antenna tilt angles, terrain models,building models, etc. These models can be continually updated over time as the UAVflies more missions. These models can also incorporate any additional informationavailable about the antennas / cells. For example, if the tilt angles of antennas in the network are available, the models could incorporate this information for higher accuracy. 2. Building a per-antenna / per-cell model and combining these into a singlecollection-of-models has advantages: this collection-of-models can predict both handoverand SINR: the two dynamics that contribute most to poor cellular UAV performance. Previous approaches which build a single model of signal strength spatial distributioncannot reason about handover between antennas / cells. Building an explicit cellularnetwork model has advantages over a model-free learned system such as in Reference [3]. A model-based system can use the same dataset, collected during varying types of UAV trajectory, to carry out trajectory planning for different types of UAV trajectory. Thisallows reuse of data from dissimilar UAV missions in the same operating area. Bycomparison, for Reference [3], if any aspect of the problem or objective were changed,the system would need retraining with new trajectories. Physical priors can beincorporated into the model, to reflect known antenna / cell properties and allow themodel to better generalize outside of the dataset coverage area. In our implementation, a Gaussian process (GP) model is used for each antenna / cell. Physical priors are provided by designing a GP kernel function, which is described in more detail in the“Probabilistic link gain models” subsection in the Detailed Description below. Anotherexample per-antenna / per-cell model is a physics-informed neural network. This type ofmodel could use the equation for signal propagation as a physical prior, to ensure thatits predictions are physically consistent and sensible. Separate antenna / cell models havelarge advantages if the antenna / cell models are probabilistic. Probabilistic models allowthe trajectory planner to determine how certain the model is about its predictions. The model will be less certain further away from the locations of its RSRP measurements.Maintaining a probabilistic model for each antenna / cell allows: the planner can avoidareas where all antenna / cell models are uncertain, as it is unsure what the SINR andhandover behavior might be in that area, the planner can properly consider theindividual aspects of uncertainty in each antenna / cell model: E.g. when reasoning abouthandover between two antennas / cells, high uncertainty in the model for a thirdantenna / cell that is further away will not have much impact on the handover dynamicsand can likely be ignored. Fig. 3A and Fig. 3B show an example output of the method according to an embodiment of the present disclosure. As can be observed from Fig. 3A, it shows a same cellular environment as that shown in Fig. 2A. However, a different trajectory than thestraight one shown in Fig. 2A is planned for the UAV this time. To be specific, the UAV ispermitted to take up to a 50% longer trajectory than the straight-line trajectory shownin Fig. 2A. When compared to the straight-line trajectory shown in Fig. 2A, this trajectoryhas 3x lower time spent disconnected (from ~12 seconds shown in Fig. 2B to ~4.1seconds shown in Fig. 3B), and is connected to the network ~96.6% of the time shownin Fig. 3B compared to ~85.3% in Fig. 2B.Next, a detailed description of an exemplary method for generating a CWM andan exemplary method for planning a trajectory will be described with reference to Fig. 5and Fig. 6, respectively. Fig. 5 is a diagram illustrating an exemplary procedure for generating a cellularworld model according to an embodiment of the present disclosure. In someembodiments, the procedure may be performed by the CWM generating device 145.The procedure may begin with step S505 where an RSRP dataset may be builtand / or updated. In some embodiments, this step may consist of adding initial RSRPmeasurements to the dataset (e.g., the initial RSRP dataset as shown). In someembodiments, this step may consist of updating the existing RSRP dataset with the newRSRP measurements (e.g., the new RSRP measurements obtained during the step S635 shown in Fig. 6). In some embodiments, this step S505 could include a filtering sub-stepthat determines whether measurements should be included in the dataset. For example,older measurements which disagree with newer measurements (e.g., differ from the newer measurements by a specified margin) may be identified, which suggests that the true underlying link gain spatial distribution may have changed. In such a case, the oldermeasurements can then be removed from the dataset.At step S510, the RSRP dataset built / updated at step S505 may be stored. Insome embodiments, the RSRP dataset may have an exemplary format as follows.The RSRP dataset ^= { (cell unique ID, RSRP measurement, ^^^^^^^^^^^^ , timestamp) , ...}consists of: cell unique ID is a unique identifier for the cell / antenna that produced thereference signal measured by the RSRP measurement; RSRP measurement is themeasurement, e.g. in dBm units. ^^^^^^^^^^^^ is the position of the UAV at the time themeasurement was taken. This could be determined using Global Positioning System(GPS) / Global Navigation Satellite System (GNSS). Timestamp is a record of the timewhen the measurement was taken. However, the present disclosure is not limited thereto. In some other embodiments, additional information (e.g. SINR measurements and / or their corresponding positions / timestamps) may be comprised in the RSRP dataset, or any of the above listed information may not be present in the RSRP dataset. At step S515, the RSRP dataset may be separated by cell / antenna ID. In someembodiments, this step may consist of separating the full dataset ^ into ^^^^^ separatedatasets ^^, each of which may only contain measurements for the cell ^, which has aunique cell ID. In some embodiments, ^^^^^ is the number of unique cell unique IDs in ^.At step S520, a cellular world model (CWM) may be generated. In some embodiments, the step S520 may comprise multiple steps / sub-steps S521, S525, and S527. For each antenna / cell ^ or each dataset ^^, where ^ = 1, … , ^ and where ^ = ^^^^^,the step S521-c is performed where a corresponding per-cell link gain model ^ may be fitto the dataset ^^. In some embodiments, this step may consist of fitting a probabilisticspatial link gain model ρ^^(^) to the cell / antenna ^, using the dataset ^^ and any otheravailable information. In some embodiments, this step could include explicitly estimatingthe location, orientation, and / or tilt angle of the antenna. In some embodiments, thisstep could include optimizing the hyper-parameters of the model to best match the data.In some embodiments, for the specific probabilistic link gain model type (a Gaussianprocess, GP) that is described in the next step S525, fitting an antenna / cell model to thedata may consist of at least one of: specifying (if known) or estimating (if unknown) thecorresponding antenna location, then optimizing the GP model hyper-parameters usinggradient descent on the log marginal likelihood. In some embodiments, the method for fitting a GP model to a per-antenna / per-cell link gain dataset differs from any existing methods at least in that: a new type ofmodel that is well-suited to predicting link gains from cellular antennas is developed.This model is never described elsewhere. In some embodiments, the model may be a model using a combination of two GPkernel functions: a polar coordinate system GP kernel and a Cartesian coordinate systemGP kernel. In some embodiments, the polar coordinate system GP kernel may use a circular distance function to generalize antenna propagation in a circular manner, centeredaround the antenna location. This type of GP kernel has not been applied to antennapropagation patterns before. In some embodiments, this component may capture large-scale (large variation in space and link gain) link gain variation caused by antenna nulls. In some embodiments, the polar coordinate system GP kernel may be implemented as follows. Large-scale variation kernel: ^Transform dataset to polar coordinates: ^^^^^^ = {(^ cos ^ , ^ sin ^), ^ ∈ ℝ, ^ ∈[0, 2^]}^ A polar kernel ^POLAR(^, ^^) = ^^ ^α ^^ ^angular(θ, θ^) + α ^^ ^^^^^^^(ρ, ρ^)^ may bedefined, o^^^^^^^(ρ, coordinatesystem radial axis, for example a common Radial Basis Function (RBF) kernel function. o^angular(θ, θ^) may be a GP kernel that applies in the polar coordinatesystem angular axis. In some embodiments, multiple options for thiskernel may be used, and they may comprise at least one of:angular , θ^ = ^^ ^2 sin ^^ ^^^ ^ (θ )^ ^. ^^is the 2^-Wendland function, and kernel is also known as the “chordal” ^^angular(θ, θ^) = ^^ ^acos^cos(^ − ^^)^^, where ^^is the ^^-Wendland function, acos(∙) is the arccosine function, and cos(∙) is the cosinefunction. This kernel is also known as the “geodesic” kernel.^ A standard GP kernel e.g. RBF.^ This can be more numerically stable than the geodesic or chordalkernels. It also has a less restricted range of lengthscale hyper- parameter values compared to the chordal / geodesic kernels. ^Note that this kernel type won’t be able to generalize aroundthe [0, 2^] boundary.^ In a specific implementation, the angle is in a real numberand not within a closed set between 0 and 2^. So there is no “looping from 2^ to 0 once complete a circle”. In short, theangle is treated as if it were a coordinate. o^^, α^^, α^^, are weighting factors (or variance factors) of this polar kernelfunction. Small-scale variation kernel: ^The Cartesian coordinate system GP kernel may have a smaller lengthscalethan the polar coordinate system GP kernel. oThis component may capture smaller-scale link gain variation caused byenvironment shadowing / slow fading. oA specific implementation may be: a Matérn kernel function, e.g., a^ Matérn ^ kernel function. In some embodiments, the two kernels (i.e., the polar coordinate system GPkernel and the Cartesian coordinate system GP kernel) may be combined by asummation, or any other appropriate operations. The present disclosure is not limited thereto. As shown in Fig. 3A and Fig. 3B, it have been demonstrated in simulation studiesthat the model accuracy of our method is improved when compared to commonly usedGP kernel functions.At step S525, probabilistic link gain models may be obtained as the outputs of thesteps S521-1 through S521-n. In some embodiments, this component may be a set oftrained / optimized probabilistic link gain models ρ^(^), one for each antenna / cell whosecell ID is represented in the dataset ^. An example implementation of a probabilisticspatial link gain model could be a Gaussian process: ρ^(^) ∼ ^^(^ ∣ ^^)That is, the probabilistic spatial link gain model ρ^(^) is distributed as a GP modelgiven the dataset ^^, or the gain model ρ^(^) is approximated bya GP process that is a function of states ^ given the dataset ^^. How the GP model isfitted for each antenna / cell is described with reference to step S521 above. As shown in Fig. 5, the procedure may further comprise step S517 where networkdata and / or assumptions may be input or otherwise obtained for step S527 where thecellular world model is built. In some embodiments, step S517 may input informationabout the additional variables required to build the cellular world model, for example, thevariables that cannot be sourced from the probabilistic link gain models obtained at stepS525. In some embodiments, the variables may include at least one of:1. Utilization factor (0.0 ≤ utilization factor ^^ ≤ 1.0) of each cell / antenna. Thismay correspond to the data demand coming from users served by that cell / antenna. 2. Variables that determine the behaviour of the handover procedure. The mostcommon / conceptually simple handover control variables may comprise at least one of the handover time-to-trigger (e.g., in seconds) and the handover hysteresis (e.g., in dB) values. In some embodiments, the input data about the variables may be supplied indifferent forms, for example, comprising at least one of (but not limited to): -Exact true knowledge of a variable’s value. For example, if the network operatoris running this system, they would be expected to know which values they have set for the handover control variables. Alternatively, values could be queried from the networkvia an Application Programming Interface (API). An assumed value for the variable,assumed to be close to the true value. A probability distribution over variable values. Aworst-case value for the variable. In some embodiments, the data provided at step S517 may influence thebehaviour of the system: 1. Providing worst-case values will likely lead to overly conservative trajectories, but this may be acceptable in practice and may not have anexcessive impact on the trajectory’s performance in the real world. For example, theworst-case value of utilization factor would be 1.0 for all cells. This is a commonly chosen value for existing communication-aware navigation works because it represents the worst-case interference that could be received, hence the highest SINR. Existingworks still achieve useful results even with a pessimistic / worst-case assumption. 2.Providing accurate values or distributions for as many variables as possible will improvethe method’s performance. For good performance, the real-world value of a variableshould be inside the support of the provided distribution for that parameter. However, if the probability distribution is too wide (too much uncertainty about the parameter value), the method may not be able to devote sufficient planning effort to the real-world value, as it must share its planning effort between many other possible real-world values ofthat parameter. If the real-world values differ too much from the values provided here,the method will not perform well and may not outperform a simple straight-linetrajectory to the goal state. These values can be learned from previous UAV trajectories.For example, handover process parameters are rarely changed and could be recorded for each cell visited during a previous trajectory. This would give exact knowledge of the parameter values for the next trajectory. At step S527, the cellular world model may be built. In some embodiments, thecellular world model may comprise at least one of: one or more per-cell link gain models(e.g., those obtained at step S525); one or more per-cell SINR models (e.g., thoseobtained in a similar manner as the one or more per-cell link gain models); one or moresampled versions of the models mentioned earlier (e.g., those described below withreference to Fig. 6); one or more variables (e.g., those obtained at step S517 and othervariable described below); and other necessary and / or optional information.In some embodiments, an exemplary definition of a cellular world (or an instanceof the cellular world model) may be provided as follows. However, the present disclosureis not limited thereto. Cellular world definitionA cellular world ^ may be a collection of parameters and / or distributions that fullydescribe a cellular environment. It may contain at least one of:^ a fixed instance of a link gain spatial distribution ^^(^) for each cell ^( ) ^ ^^(^)^^^^^o Let ^ ^ = ^^ ^^^be the function that gives the vector of the link gain to all cells / antennas at position ^.^ a fixed instance of an SINR spatial distribution ^^(^) that gives the expected (i.e.expected time-averaged, the true SINR will be constantly varying in time) SINR that the UAV would experience at position ^ while associated with cell ^.o Let ^(^) = ^^^(^)^^^^^^ ^^^be the function that, given a position ^, gives the vector of expected SINR values that would be experienced by the UAV for each cell ^ itcould be associated to.^ all the variable values required by the cellular transition system (which will bedescribed in details with reference to step S611 shown in Fig. 6): ofixed values of handover and / or utilization variables,o fixed values for the successful / failed handover interruption times. Alternatively,if the successful and failed handover interruption times are random variables, the parameters for these random variable distributions.^ and fixed values for any other pertinent variables.There exists a ground-truth cellular world ^GT, from which the RSRP dataset is collectedand in which the UAV will fly its trajectory. What the ground-truth cellular world is unknown. At step S530, the cellular world model may be obtained as the output of the stepS527. In some embodiments, the cellular world model may be a model that representsthe system’s belief about the ground-truth cellular world. In some embodiments, it maymaintain the probability ^(^^) that ^^ is the ground-truth cellular world, for any ^^within the support of ^(^^). In some embodiments, the cellular world model may bequeried by the trajectory planning (e.g., the system described with step S610 shown in Fig. 6) to evaluate the performance of UAV trajectories in different possible cellular worlds. With the procedure described above, a cellular world model may be created,updated or otherwise maintained. Although only the fitting of the per-cell link gainmodels is described during the step of S520, the present disclosure is not limited thereto.For example, when SINR measurements can be obtained, one or more per-cell SINR models may be designed and fitted to corresponding per-cell datasets, and become parts of the cellular world model, in addition to or as an alternative to the one or more per-cell link gain models. Next, a detailed description of planning a trajectory for a UAV based on a cellular world model may be given with reference to Fig. 6. Fig. 6 is a diagram illustrating an exemplary procedure for planning a trajectory for a UAV based on a cellular world model according to an embodiment of the presentdisclosure. The procedure may begin with step S605 where one or more cellular worldsmay be predicted or queried from the cellular world model (e.g., the cellular world modelobtained at step S530 shown in Fig. 5). In some embodiments, the cellular world model could provide its predictions indifferent forms: It could produce samples of possible true cellular environments^^^^^^^ ~ ^(^) when queried. The embodiment described below uses this method. At ahigh level, many possible ground-truth cellular worlds may be sampled from the cellularworld model and trajectories that can perform well in these sampled environments maybe generated. Additionally, in some embodiments, the method can react during the real-world trajectory as it determines which of its cellular world samples are the most similarto the ground-truth cellular world it has experienced so far during the trajectory. Insome embodiments, to sample a valid cellular world, the sampled spatial link gain spatialdistribution, the sampled SINR spatial distribution, and other sampled variable valuesmust be consistent. For example, when SINR ^(^) = ^(^(^), … ) is a function of link gain^(^) , ^(^) and ^(^) are consistent to each other when ^(^) = ^(^(^), … ) . Given acellular world ^^^^^^, it could provide the probability that ^^^^^^ = ^^^. It could givethe probability of the true cellular environment lying within a given space of cellular worlds ^^^^^^. In some embodiments, a sampling-based implementation for generating ^(^) maybe provided as follows: Step S605 may be used by the trajectory plan generation systemto sample a possible cellular world from the cellular world model. For a collection of GPlink gain models (or in general, a collection of GP models), a memory- and time-efficientmanner of sampling worlds is to: when first initialising the system: 1. Draw ^^^^^^^samples from each GP model ^^^^^^^^^, evaluated at a grid of points ^^^^^that coversthe operating area, which could be specified at step S601. 2. Store these ^^^^^^^samples. The original GP prediction object (e.g., the original GP model itself) does notneed to be stored, and the size of this object is much larger than the size of the samples.When queried for a new cellular world sample: 1. Choose a random index 0 ≤ ^^^^ <^^^^^^^ for each cell antenna / cell ^. 2. For each cell / antenna ^’s sample collection, takethe sample at index ^^^^. 3. Use this collection of samples to create linear interpolators720 between the points ^^^^^. These linear interpolators 720 can be used together as thesampled function ^(^).An exemplary sampling procedure is shown in Fig. 7. As shown in Fig. 7, multiplesamples may be drawn from each of the per-cell link gain models 700-1 through 700-n,and a sample may be selected randomly from the samples drawn from each per-cell linkgain model. For example, for the per-cell link gain model 700-1, multiple samples are drawn, and one of the samples (e.g., the sample 710) is selected. Similarly, for each of other per-cell link gain models 700-2 through 700-n, a sample is randomly selected, andthese selected samples can be used to create linear interpolators 720, one for each cell,which can be used together as the sampled function ^(^) that can provide an input (e.g.,a link gain vector comprising link gain values) to a cellular transition system 800 asshown in Fig. 8. For example, for the cell corresponding to the sample 710, an interpolator may becreated from the sample 710 as follows. In some embodiments, for any position ^ on thegrid ^^^^^, the interpolator may output a link gain value ^^,^ that is exactly equal to thelink value of the sample 710 at the same position ^. For any position ^ off the grid ^^^^^(e.g., a position between two or more on-grid positions), the interpolator may output alink gain value ^^,^ that is a linear combination of the link gain values of the sample 710at the two or more on-grid positions. Similarly, other interpolators created for other cellsmay output link gain values ^^,^, …, ^^,^ for the same position ^, respectively, and theselink gain values including ^ may form a l^ ^,^ ink gain vector ^^ = ^^^,^, … , ^^,^^for thisposition ^ , which may be input to the cellular transition system 800. In someembodiments, the position ^ could be the position ^^ of the UAV at the time instance ^^and the link gain vector ^^could be the link gains for all the cells at the position ^^,where ^ = 1, … , ^, as will be described with reference to Fig. 8.Further, there are many other ways to create the sampled function ^(^), and thepresent is not limited to the linear interpolator method described above.This is memory-efficient because only a small number of samples are stored perlink gain model, but the number of possible ^(^) functions that can be sampled is^^^^^^^^^^^^, which is very large even with a small number (~100+) of samples. This is time-efficient because once the ^^^^^^^samples have been generated atthe initialization time, sampling a new instance of ^(^) is fast. The linear interpolatorspatial mesh can be pre-calculated and reused, because all samples are at the same points ^^^^^. Further, although a grid of points ^^^^^may be used as the positions where the samping is done in the above embodiment, the present disclosure is not limited thereto. In some other embodiments, the sampling positions may comprise position points that do not form a grid. In general, the sampling positions may comprise a set of one or more position points, which can be determined arbitrarily and / or as required, for example, separated from each other by a same distance or different distances. Further, although the step S605 is described as a part of the method, it may be omitted, for example, when the original per-cell models are used instead of the sampled per-cell models, or when the sampled per-cell models are provided as inputs. Further, in some other embodiments, the per-cell models may comprise original per-cell models, sampled per-cell models, or a combination thereof. Next, an exemplary sampling-based implementation to generate ^(^) will bedescribed below. In some embodiments, the expected SINR (discounting background noise, whichis small compared to interference for the UAV downlink domain) may be defined asfollows: γ( )ρ^(^) ^^ = ^ ^∑^ ^^ ^ ^ ^^^^ ∙ ρ^^(^)^where ^^^ is a binary variable specifying whether ^ is transmitting ata instant in time: ^^ ∼ Bernoulli(^^) where ^^ is the factor for cell / antenna ^.In some embodiments, ^^ may be specified by the input data / assumptions (e.g.,that at step S517 shown in Fig. 5). If the method is given a distribution over ^^, a value^^ may be sampled and used here instead.In some embodiments, due to the difficulty of calculating this expectation, a lower-bound on the SINR may be derived using Jensen’s inequality. For the link gainspatial distribution functions ^^^(^) sampled as described above, a sampled lower-boundSINR ^^(^) may be given by:^^ ^ ^ ^^^ (^) ≥( )^^∑^^^ ^ ^^^^ ∙ ^^^(^)^ ^ ^ ^^(^) = ^( )∑^^^ ^ ^^^^ ∙ ^^^^(^)^ For computational efficiency reasons, rather than pre-compute a linearinterpolator for SINR across the whole domain, only this value is computed when thetrajectory sampler requests the value of ^^(^) at a point ^.^(^) is therefore a sampled lower-bound on SINR. This is fine because we careabout ensuring a minimum SINR: if we ensure the lower-bound is above the minimum, the true SINR should be above the minimum also. Further, there will be alternative methods of generating ^(^). For example, a time-sampling approach could be used which averages over multiple samples from ^^^. Although the SINR is sampled indirectly via the sampled link gain values as described above, the present disclosure is not limited thereto. For example, in some other embodiments, the SINR can be sampled in a similar manner as the link gain, for example, when the SINR measurements can be obtained directly from the UAV. In such a case, per-cell SINR models can be obtained by fitting to collected per-cell SINRdatasets, and then sampled directly.In some embodiments, any other variables provided to the cellular world by step S517 or any other source can also be sampled. For example, if a distribution over handover time-to-trigger has been provided, a value for handover time-to-trigger may be sampled from that distribution to apply in this cellular world. At step S601, an operating area and / or an operating space may be input. In someembodiments, the operating area and / or the operating space may be user-specifiedinputs, defining the UAV operating area and / or the UAV operating space, respectively. Insome embodiments, these inputs may specify a 2D polygon and / or a 3D polyhedronbounding the area and / or the space the drone is allowed to operate in, respectively. Insome embodiments, the start and goal locations should be within this area and / or space.In some embodiments, with the current implementation, the 2D bounding polygonand / or the 3D bounding polyhedron may be required for the method to work because itdefines ^^^^^. In some embodiments, with changes to the current implementation (e.g.calculating ^^^^^ using the space of possible trajectories defined by step S602 instead ofusing the operating area polygon / the operating space polyhedron), this could be anoptional field that does not need to be provided if the UAV has no specified operatingarea / space. In some embodiments, this input could optionally include a 2D or 3D mapspecifying areas / spaces that the UAV should not enter, in order to avoid obstacles or no-fly areas. At step S602, one or more trajectory requirements may be input. In someembodiments, this may be a user-specified input, specifying the physical and / orcommunication requirements for the output trajectories. In some embodiments, thetrajectory requirements specification may include (but not limited to) at least one of: Thedesired UAV operating altitude: the current implementation may generate trajectories in 2D at the specified operating altitude. It does not allow the UAV to alter its trajectory inthe altitude dimension. However, the present disclosure is not limited thereto. In someother embodiments, a trajectory involving different altitudes may be planned; The startposition ^^^^^^; the goal position ^^^^^, or some other specification that the UAV trajectoryshould satisfy, for example, a more complex specification could be a set of locations thatthe UAV should visit at specified times over the course of the mission; SINRmin, the minimum required SINR for the UAV to be connected to the network. This could be set based on the minimum required bitrate to the UAV required to reliably control the UAV; specification on the allowable disconnection time for the trajectory. Different specifications / objectives are discussed above with reference to step S527; the maximumtime allowable for the trajectory to reach the goal ^^^^ . In some embodiments, thiscould be provided as a relative multiplier: e.g. 1.5X means that the trajectory can take up to 50% more time than the shortest possible (straight-line trajectory) time the UAV could reach the goal from the starting point. At step S610, a trajectory generation algorithm that uses the cellular world modelto generate UAV trajectories may be performed given requirements specified in stepsS601 and S602. In some embodiments, a cellular transition system(s) may be used topredict the communications quality along the trajectories being evaluated by thealgorithm. In some embodiments, the step S610 may comprise multiple steps / sub-stepsS611, S613, and S615. At step S611, the cellular transition system may be initialized. In someembodiments, the cellular transition system may be a stateful object which may trackthe cellular state along a simulated UAV trajectory. In some embodiments, a cellulartransition system can be initialized given a cellular world instance (e.g., the instanceobtained at step S605). In some embodiments, the behaviour of the transition systemmay be parameterized by values contained in the cellular world, for example, thehandover time-to-trigger and hysteresis values. In some embodiments, the cellular transition system could be implemented indifferent ways: It could act as a simulator that simulates the cell antenna association andhandover process at discrete, successive moments in time, given inputted RSRP and SINR values over time (“black-box simulator” type). This is the type that will bedescribed below. It could give explicit probabilities of communication quality andtransitions between cell associations, for a specified cellular state and time period. Fig. 8 is a diagram illustrating an exemplary cellular transition system 800according to an embodiment of the present disclosure. Fig. 8 shows a “black-boxsimulator” type cellular transition system 800. In some embodiments, it may beinitialized with parameters from the cellular world. For example, given equal-length vectors of time values, SINR vector values, andlink gain vector values, the cellular transition system 800 may simulate the evolution ofthe cellular network connection state. In some embodiments, it may output the cell that the simulated cellular connection is associated with, and incurred disconnection times (due to handover or insufficiently high signal strength) at each time point specified bythe time value input vector. In some embodiments, it may be stateful, so when it is nextrun with 3 new vectors, it may continue from the state it was in at the end of theprevious set of vectors.In some embodiments, the cellular transition system 800 may use the simplestcase of handover behaviour: Handover occurs when a cell / antenna has a higher link gainvalue than the current associated cell link gain value by more than <handover hysteresis>dB, for at least <handover time-to-trigger> seconds. Radio link failure (causing a longdisconnection time) occurs when the SINR of the currently associated cell falls below a minimum level. However, the present disclosure is not limited thereto. In some otherembodiments, more complex and proprietary handover behaviours could be simulated ifdesired and given sufficient knowledge of the handover behaviour in the true network.Referring back to Fig. 6, at step S613, a UAV trajectory plan may be generated. Insome embodiments, this can be done by planning control actions. The remainder of thisstep may consist of implementation-specific details. However, the present disclosure isnot limited thereto. In some other embodiments, other possible implementations of atrajectory planner may be used.In some embodiments, a trajectory planner implementation may minimize theexpected total disconnection time given a maximum time budget ^^^^to reach the goalstate. A single UAV trajectory action is (^, θ, t). This instructs the UAV to set its targetvelocity to ^ and to travel at an steering angle θ for time ^. A fixed UAV trajectory is alist of UAV trajectory actions [(^^, θ^, t^), (^^, θ^ , t^).. , (^^ , θ^, t^)]. A reactive trajectory isdependent on the UAV’s state. The definition of the UAV state is provided below. In theembodiments described below, reactive trajectories are produced.In some embodiments, a state may be formed of observable and unobservablecomponents. Observable components are variables that are directly available to the UAV.For example, it can be assumed that its position ^ is observable via GPS / GNSS.Unobservable components are variables that cannot be directly observed by the UAV, but it may be possible to infer their values by analyzing their effects on the UAV. For example, the full state of the handover process is not observable to the UAV, because handover is managed by the cellular network rather than by the UAV. In some embodiments, the UAV observable state components may be: ^^ ∈ ^ × Θ × ^ × ^Physical state: ^ is the 2D position of the UAV. θ is its heading angle. These are inthe coordinate system of the operating area. ^ is the trajectory time value. The start ofthe trajectory is at ^ = 0. ^ is the cell antenna the UAV is associated with.It would be possible to include more available cellular variables in the observable state, such as the latest values of RSRP measurements for the current associated cell, or multiple cells. This would enable the UAV to make decisions which make use of moreinformation, but a larger state space makes the trajectory planner’s job more difficult.In some embodiments, the UAV unobservable state component may be theinternal state λ of the cellular transition system: ^^ ∈ Λ.A UAV reactive policy ^(^) is a mapping ^^ → ^ × Θ × ^ which maps UAV states tothe action that should be taken in that state. One way to make decisions when part ofthe state space is unobservable is to instead use a history-dependent policy π(ℎ) whichis a function of history ℎ = ^^^^^^^^ … ^^^^^^ (where ^^ refers to a state and ^^ refers toan action) rather than solely the current observable state ^^: ^ → ^ × Θ × ^.In some reactive history-dependent policies may be generatedusing an off-the-shelf Bayesian RL algorithm. This is an ideal class of algorithms for decision-making in the face of model and state uncertainty, as we have in our problemsetting. There is model uncertainty, because the UAV does not know the true cellularworld that it will fly its trajectory in. There is state uncertainty, because the UAV doesnot know the values of unobserved state variables. Next, a high level description of the planning algorithm will be provided below.The algorithm used in this embodiment is a simulation-based Monte Carlo TreeSearch (MCTS) algorithm: Katt, S., Oliehoek, F.A. and Amato, C., 2017, July. Learning in POMDPs with Monte Carlo tree search. In International Conference on Machine Learning (pp. 1819-1827). PMLR., which is incorporated herein by reference in its entirety. An MCTS algorithm may generate a policy by simulating a finite number of trials. Using these trials, it may build a search tree. A full description of the algorithm’s operation is beyond the scope of the present disclosure, but a few key points that definethe specifics of how MCTS is applied to our problem instance will be given below.Fig. 9 is a diagram illustrating an exemplary Monte Carlo Tree Search (MCTS) search tree for trajectory planning according to an embodiment of the present disclosure.Fig. 9 shows where in the search tree different steps take place. For example, it showsfull states ^ being stored in each search tree node: in reality, each search tree nodecontains only observable state components (“observations”) and full states aremaintained by a particle filter. In some embodiments, each trial may start from the root node, which is the initialstate. In some embodiments, at each state visited during the trial, an action (in our case,a UAV trajectory action) may be selected according to some formula that balances tryingnew actions and trying actions that have proven to be beneficial in past trials. In someembodiments, when an action is taken, the outcomes of that action may be simulatedstarting from that state. The outcome may be a new state. In some embodiments, if thenew state is already represented in the search tree (e.g. action ^^from state ^^resultingin ^^ in Fig. 9), the trial may continue from that state. In some embodiments, if the newstate is not already represented in the search tree (e.g. action ^^ from ^^ resultingin ^^ in Fig. 9), it may be added to the search tree and its cost-to-goal is estimated byperforming a rollout from that new state. In our case, a trial may finish when it reaches the goal point state or performs arollout. The trial may have incurred some total cost ^^^^^^, and this total cost may beused to update the value of each search tree node that the trial passed through. In someembodiments, search tree node values may be used by action selection in future trials,and used output the final policy.On top of standard MCTS, the algorithm used here may address model and stateuncertainty by sampling a single fixed state and model dynamics at the start of eachMCTS trial, and using it throughout the MCTS trial. For our setting, sampling a fixedmodel dynamics may be achieved by sampling a cellular world (e.g., at step S605).Sampling the initial state includes initializing a new cellular transition system (e.g., atstep S611), whose internal state may be the initial unobservable state. In someembodiments, a new cellular transition system may be initialized with a random initialstate or a belief-informed initial state. In real life, some parts of the internal state of acellular modem are not observable by the user. For example, the values of internaltimers governing e.g. handover cannot usually be queried via an API. In our case, the"user" is an algorithm trying to optimise communications performance. Even though thealgorithm does not know the internal state, the algorithm can have beliefs orassumptions about the internal state. For example, if a timer only measures timebetween 0 < t ≤ 3 seconds, the algorithm can have a uniform belief about the timervalue, between 0 and 3 seconds. That is, the algorithm believes any timer value between0 and 3 seconds is equally probable. The class of algorithm use herein may operate bysampling possible internal states from this belief. The algorithm may plan by simulating many possible "episodes". At the start of an episode, the algorithm would sample a"true" value for the timer from a uniform distribution 0 < t ≤ 3 seconds, and initialise thecellular system using this value. When executing, even though the UAV does not observethe unobservable state, its trajectory should still be robust to any possible initial statevalues. The sampled cellular world may inform the trajectory planner of how the stateshould evolve after each action. The process will be described in the next steps.Fig. 10 is a diagram illustrating an exemplary simulation of a motion action between states according to an embodiment of the present disclosure. In someembodiments, an action simulation may return a new state and a corresponding cost forthe action. As shown in Fig. 10, an exemplary simulation of an exemplary motion actionfrom state ^^ resulting in state ^^ is performed. The curved line shows the trajectoryfollowed by the UAV given the action velocity / steering / angle / total time, and the pointsindicated by the symbols “X” are points along that trajectory at evenly spaced timevalues ^^…^. The corresponding positions at these time values are ^^…^.To simulate an action (^, θ, t) starting from a state ^^ = (^^, θ^, ^^,∙ ), the trajectoryproduce a sequence of ^ trajectory points ^^^^^ = [^^, ... , ^^] using the variables (^, θ, t) specified by that action. This is illustrated in Fig. 10, for ^ = 5 .Trajectory points may be separated by a constant timestep starting at time ^^, and givethe position of the simulated UAV after time ^^. Trajectory points may correspond to timevalues [^^, ... , ^^].In some embodiments, the cellular sampled for this MCTS trial may be usedto predict a sequence of link gain vectors [^(^^), … , ^(^^)] and a sequence of SINR valuevectors [^(^^), … , ^(^^)], at the trajectory points.The cellular transition system 800, starting from its internal state at state ^^, nowprocesses the sequences of time [^^, … , ^^], link gain vectors [^(^^), … , ^(^^)] and SINRvalue vectors [^(^^), … , ^(^^)] for the trajectory. Its internal state evolves, and itoutputs a sequence of cell … , ^^] and a sequence of disconnection times[^^, … , ^^].The observable component of the new state is ^^ = (^^ , ^^ , ^^ , ^^) . Theunobservable component of the new state is the new internal state of the cellulartransition system 800.The cost for this simulated action is the total disconnection time experiencedduring the trajectory: ^ = ∑^ ^^^ δ^. Fig. 11 is a diagram illustrating an exemplary rollout of a trajectory to a goal position from a state according to an embodiment of the present disclosure. In someembodiments, a rollout simulation may return incurred cost to reach goal from a statewhen doing rollout. As show by Fig. 11, an exemplary trajectory may be rolled out to thegoal point from a state ^ . In some embodiments, a rollout trajectory may be theminimum time trajectory that reaches the goal point starting from the state position ^and angle θ.In some embodiments, to carry out a rollout from a state ^ = (^, ^, ^,∙ ) to the goalpoint ^^^^^ , a shortest-path trajectory may be generated between ^^^^^^ and ^^^^^ . Insome embodiments, this path may be generated analytically as a Dubins path using theminimum turning circle of the UAV at its top speed. In some embodiments, the UAV mayfollow the generated path at its maximum speed until it reaches the goal point. In someembodiments, the cost incurred along the rollout trajectory to the goal position may becalculated in the same manner as in action simulation. Because Dubins path generationis analytic, calculating the rollout (minimum) trajectory length from any position and angle is fast. Enabled actions In some embodiments, the available actions at each search tree state may berestricted, in order to ensure that the trajectory must be complete within the timebudget. With the action definition described above, each action may have a deterministicend position and angle given the state it is taken from. Let ^^, θ^be the new positionand heading values after taking action (^^^^, ^^^^ , ^^^^) from the current state ^ =(^, ^, ^,∙ ), and let ^^^^^^^^(^, ^) be the time taken to do a rollout trajectory from position ^and heading ^. In some embodiments, an action (^^^^, ^^^^ , ^^^^) may be enabled in ^ = (^, ^, ^,∙ )if ^ + ^^^^ + ^^^^^^^^(^′, ^′) ≤ ^^^^.In some embodiments, an action may only be enabled if taking that action stillleaves enough time to reach the goal state from the new position / heading, without exceeding the time budget. In some embodiments, if no actions are enabled, only the rollout trajectory can be taken from this state. Online MCTS In some embodiments, some algorithm implementations, such as theimplementation described above, are able to carry out additional trajectory optimizationcomputation online. This may enable the UAV to better react online, rather than relyingon reacting based purely on pre-computation of the trajectory policy.At step S615, whether the trajectory meets requirements or not may bedetermined. In some embodiments, this component of the trajectory generation blockmay evaluate whether the trajectory planner’s output meets the trajectory requirements.It may be that the input trajectory requirements cannot be satisfied, given e.g. too stringent trajectory requirements or too high uncertainty over the cellular environment.In such a case, step S620 may be performed where inability to find valid solution may bereported. In some embodiments, if the generated trajectory does not meet therequirements, the trajectory planner may return control to the user / caller and indicatethat it could not generate a solution within the provided trajectory constraints, andoptionally its level of knowledge of the cellular environment (e.g., when the failure iscaused by lack of knowledge of the cellular environment).At step S625, whether UTM reporting is required may be determined. In someareas / jurisdictions, reporting UAV trajectory plans to an unmanned traffic management(UTM) system is necessary. In such a case, step S630 may be performed where theplanned trajectory may be reported, for example, to the UTM 150. In someembodiments, if the trajectory plan is reactive, such as in the implementation describedabove, then the system cannot report a single fixed trajectory to the UTM system. Insuch a case, some options for UTM integration may be then: the trajectory planningsystem could request a space / time reservation that covers all possible trajectories thatthe UAV might take. The online trajectory planning system could request authorizationsfor trajectory deviations during flight, if the UTM system is able to respond quickly enough. At step S630, if the UTM 150 rejects the proposed trajectory, control flow mayreturn to the trajectory generator to generate a new trajectory option (e.g., step S610).At step S635, one or more operations may be performed, including (but notlimited to) at least one of: execute the UAV trajectory, authorize reactive deviations fromoriginal path plan, record RSRP measurements, carry out additional trajectoryoptimization online. In some embodiments, step S635 is the final step of a single trajectorygeneration / execution cycle. In some embodiments, the trajectory may be sent to theUAV, and the UAV may execute the trajectory. In some embodiments, the UAV mayrequest permission for trajectory deviations from UTM during the trajectory. In someembodiments, while executing the trajectory, the UAV may record RSRP measurements. In some embodiments, after the end of the mission, or during the mission, the UAV maysend new RSRP measurements to be added / updated to the RSRP dataset (e.g., asdescribed with reference to step S505 shown in Fig. 5). In some embodiments, most of the components shown in Fig. 5 and Fig. 6,excluding components which must take place on the UAV (e.g. “executing the UAVtrajectory plan on the UAV”), could run in a distributed manner, e.g., on a cloud. Insome embodiments, cellular world model generation / maintenance and trajectoryplanning may be separated into two separate nodes (e.g., the CWM generating device145 and the trajectory planning device 140). In some embodiments, cellular world modelgeneration / maintenance may benefit from parallel hardware such as GPUs, and trajectory planning may be more CPU-intensive. Fig. 12 is a flow chart illustrating an exemplary method 1200 for generating acellular world model according to an embodiment of the present disclosure. The method1200 may be performed at a device (e.g., the CWM generating device 145). The method1200 may comprise a step S1210. However, the present disclosure is not limited thereto.In some other embodiments, the method 1200 may comprise more steps, differentsteps, or any combination thereof. Further the steps of the method 1200 may beperformed in a different order than that described herein when multiple steps areinvolved. Further, in some embodiments, a step in the method 1200 may be split intomultiple sub-steps and performed by different entities, and / or multiple steps in themethod 1200 may be combined into a single step.The method 1200 may begin at step S1210 where a cellular world modelindicating a belief about what a true cellular environment is may be generated.In some embodiments, the cellular world model may be able to be used forplanning a trajectory for a communication device in the true cellular environment. Insome embodiments, the cellular world model may comprise at least one of: one or moreper-cell link gain models; one or more sampled per-cell link gain models; one or more per-cell Signal-to-Interference-plus-Noise Ratio (SINR) models; one or more sampled per-cell SINR models; and one or more variables. In some embodiments, the one or more per-cell link gain models, the one or more per-cell SINR models, and their sampled versions may be spatial distribution models. In some embodiments, the one or more per-cell link gain models and / or the one or more per-cell SINR models may be probabilistic models that are fit to one or more per-cell datasets associated with corresponding cells, respectively. In some embodiments, a probabilistic model may be fit to a per-cell dataset by: optimizing one or more hyper parameters of the probabilistic model by using a gradient descent algorithm on the log marginal likelihood. In some embodiments, a per-cell dataset associated with a cell may comprise one or more entries, each of which comprises at least one of: an identifier (ID) of the cell; a Reference Signal Received Power (RSRP) measurement for the cell; anSINR measurement for the cell; a position where the RSRP measurement and / or theSINR measurement is taken; and a timestamp when the RSRP measurement and / or theSINR measurement is taken.In some embodiments, an entry in a per-cell dataset may be removed from the per-cell dataset when the entry has a measurement that differs from anothermeasurement by a specified margin. In some embodiments, the measurement may betaken at a first position. In some embodiments, the other measurement may be taken at the first position or a second position that is separated from the first position by aspecified distance or shorter, but later than the measurement. In some embodiments, atleast one entry in a per-cell dataset may have a measurement that is a weighted average of at least one of: measurements that are taken at different times; and measurements that are taken at different positions that are separated from each other by a specified distance or shorter. In some embodiments, before a probabilistic model is fit to a per-cell dataset, one or more entries in the per-cell dataset may be converted by: determining the position of an antenna associated with the corresponding cell; and converting one or more coordinates of one or more positions in the one or more entries into coordinates in a coordinate system with its origin being the position of the antenna. In some embodiments, the probabilistic model may be a Gaussian process (GP) model having a kernel function that is either a single kernel function or a combination of multiple kernel functions. In some embodiments, the multiple kernel functions may comprise at least a first kernel function based on a polar coordinate system and a second kernel function based on a Cartesian coordinate system. In some embodiments, the firstkernel function may be determined by: ^POLAR(^, ^^) = ^^ ^α ^^ ^angular(θ, θ^) + α ^^ ^^^^^^^(ρ, ρ^)^ where ^ ^ kernel functionapplies in the (∙,∙) is a GPfunction that applies in the radial axis of the polar coordinate system, θ and θ^ areangular coordinates of positions ^ and ^^ in the polar coordinate system, respectively, ρand ρ^ are radial coordinates of ^ and ^^ in the polar coordinate system, respectively,and ^^, α^^, α^^, are weighting factors of the first kernel function. In some embodiments, ^^^^^^^(∙,∙) may be a Radial Basis Function (RBF). In somembodiments, ^ ∙,∙ may be one of: ^ θ ^^ ^e ( ) ( ^) ^angular angular , θ = ^^ ^2 sin^^ , where ^^ is^ ^ ^ − ,^ - ∙function, and cos(∙) is the cosine function; and an RBF.In some embodiments, the second kernel function may have a length-scaleparameter with a value less than that of the first kernel function. In some embodiments, the second kernel function may be a Matérn kernel function. In some embodiments, atleast one of the multiple per-cell SINR models may be determined by: γ(x) = ^ρ^(x) ^^∑ ^ ( )^ ^^ ^ ^ ^^^ ∙ ρ^^ x ^ where γ^ (∙) is the per-cell SINR model for the cell ^, ^ is a position at which theSINR for the cell ^ is to be determined,where ρ^(∙) is the per-cell link gain model for the cell ^, ρ^^(∙) is the per-cell linkgain model for the cell c′ that is different from the cell ^, ^^^ is a binary variable subjectto a Bernoulli distribution with a probability of ^^^ for its value “1” and a probability of1 − ^^^ for its value “0”, and ^^^ is the utilization factor for the cell c′,where ∑^^ ^ ^ (∙) is a summation function for all cells c′ that are not the cell ^,where ^[∙] is a function for calculating the mathematical expectation of itsoperand. In some embodiments, the one or more variables may comprise at least one of: a utilization factor for at least one cell; and one or more variables associated with a handover procedure. In some embodiments, the one or more variables associated with a handover procedure may comprise at least one of: a handover time-to-trigger; ahandover hysteresis; a successful handover interruption time; a failed handoverinterruption time; one or more parameters for a probability distribution of a successful handover interruption time; and one or more parameters for a probability distribution ofa failed handover interruption time. In some embodiments, the one or more variablesmay comprise at least one of: a variable that has a known true value; a variable that has an assumed value; a variable that has a probability distribution over possible values; and a variable that has a worst-case value. In some embodiments, a sampled per-cell link gain model associated with a cellmay be an instance of a possible true underlying link gain spatial distribution. In someembodiments, the instance may be drawn from a corresponding per-cell link gain model associated with the same cell. In some embodiments, the instance may be drawn at aset of one or more position points. In some embodiments, the set of position points maycover an operating area in which a trajectory is to be planned. In some embodiments, afirst number of sampled per-cell link gain models associated with a cell may be storedand / or used in place of the corresponding per-cell link gain model associated with the same cell. In some embodiments, a sampled per-cell SINR model may be determined by: ^^(x) =^^^(x) ∑^^ ^ ^ ^^^^ ∙ ^^^(x)^where ^^(∙) is the for the cell ^, x is a position atwhich the SINR for the cell ^ where ^^^ (∙) is a sampled per-cell link gain model for the cell ^, ^^^^(∙) is a sampledper-cell link gain model for the cell c′ that is different from the cell ^ , and ^^^ is theutilization factor for the cell c′, where ∑^^ ^ ^ (∙) is a summation function for all cells c′ that are different from thecell ^. In some embodiments, the one or more per-cell datasets may be continually updated over time as more measurements are obtained. In some embodiments, the communication device may be an Unmanned Aerial Vehicle (UAV). Fig. 13 is a flow chart illustrating an exemplary method 1300 for planning atrajectory for a communication device according to an embodiment of the presentdisclosure. The method 1300 may be performed at a device (e.g., the trajectory planningdevice 140). The method 1300 may comprise a step S1310. However, the presentdisclosure is not limited thereto. In some other embodiments, the method 1300 maycomprise more steps, different steps, or any combination thereof. Further the steps ofthe method 1300 may be performed in a different order than that described herein whenmultiple steps are involved. Further, in some embodiments, a step in the method 1300 may be split into multiple sub-steps and performed by different entities, and / or multiplesteps in the method 1300 may be combined into a single step.The method 1300 may begin at step S1310 where a trajectory for thecommunication device in a cellular environment may be planned based on at least acellular world model that indicates a belief about what the cellular environment is.In some embodiments, the cellular world model may be generated according toany of the methods of the first aspect. In some embodiments, the trajectory may be one of: a fixed trajectory that cannot be changed during its execution; and a reactive trajectory that can be changed during its execution. In some embodiments, before and / or during the step of planning the trajectory,the method 1300 may further comprise: drawing, for each cell, a specified number ofsamples from a corresponding per-cell link gain model. In some embodiments, before and / or during the step of planning the trajectory, the method 1300 may further comprise: determining a sample index for each cell randomly; selecting the sample at the sampleindex from the specified number of samples drawn for each cell.In some embodiments, before and / or during the step of planning the trajectory,the method 1300 may further comprise: determining a sampled per-cell SINR model by:^^( ^^x) = ^(x)∑^^ ^ ^ ^^^^ ∙ ^^^(x)^where ^^(∙) is the for the cell ^, x is a position atwhich the SINR for the cell ^ where ^^ (∙) is the sampled per-cell link gain model for the cell ^ , ^^^(∙) is thesampled per-cell link gain model for the cell c′ that is different from the cell ^, and ^^^ isthe utilization factor for the cell c′, where ∑^^ ^ ^ (∙) is a summation function for all cells c′ that are different from thecell ^. In some embodiments, before and / or during the step of planning the trajectory, the method 1300 may further comprise: drawing a sample from a variable in the cellular world model according to its probability distribution when the variable is a random variable. In some embodiments, before the step of planning the trajectory, the method 1300 may further comprise obtaining at least one of: an operating area and / or an operating space in which the trajectory for the communication device is to be planned;one or more indicators, each of which indicates an area and / or space that thecommunication device is not allowed to enter; and one or more requirements on thetrajectory to be planned. In some embodiments, the operating area may be a two-dimensional (2D) polygon bounding an area in which the communication device isallowed to operate. In some embodiments, the operating space may be a three-dimensional (3D) polyhedron bounding a space in which the communication device isallowed to operate. In some embodiments, the one or more requirements may comprise at least oneof: a desired operating altitude for the communication device; a start position; a goalposition; one or more positions that the communication device is to visit at specified times; a minimum SINR for the communication device to maintain its connection to the network; a maximum allowable disconnection time; a maximum allowable ratio of the total disconnection time to the total time for trajectory execution; a minimum probability that the total disconnection time never exceeds a maximum disconnection time; a maximum allowable cost that is calculated based on a combination of the total time fortrajectory execution and the total disconnection time; and a maximum allowable time forthe communication device to reach the goal position. In some embodiments, before and / or during the step of planning the trajectory,the method 1300 may further comprise: initializing a cellular transition system that is able to track a cellular state of the communication device along a trajectory. In some embodiments, the cellular transition system may be initialized with a cellular world instance sampled from the cellular world model. In some embodiments, the cellular transition system may be able to determine the cellular state of the communication device along the trajectory based on one or more inputs comprising at least one of: one or more indicators, indicating a series of time when the communication device is travelling along the trajectory; one or more indicators, each of which indicates an SINR vector comprising one or more SINR values for one or more cells sampled at a position along the trajectory; one or more indicators, each of which indicates a link gain vectorcomprising one or more link gain values for one or more cells sampled at a positionalong the trajectory; and one or more cellular world parameters. In some embodiments, the one or more SINR values comprised in the SINR vector may be obtained by querying at least one of: one or more corresponding per-cell SINR models; and one or more corresponding sampled per-cell SINR models. In some embodiments, the one or more link gain values comprised in the link gain vector may be obtained by querying at least one of: one or more corresponding per-cell link gain models; and one or more corresponding sampled per-cell link gain models. In some embodiments, the cellular state of the communication device along the trajectory may comprise at least one of: one or more indicators, each of which indicates a cell, with which the communication device is associated at a specified time; and one or more indicators, each of which indicates a disconnection time during a specified period. In some embodiments, the cellular state may be determined based on at least one of one or more handovers and one or more radio link failures (RLFs). In some embodiments, a handover from a first cell to a second cell for the communication device may be determined to occur when the second cell has a higher link gain than that of the first cell by more than a threshold value, which is indicated by a handover hysteresis variable, for at least a time period, which is indicated by a handover time-to-triggervariable. In some embodiments, an RLF may be determined to occur when an SINR of acurrently associated cell falls below a minimum level. In some embodiments, the step of planning the trajectory may comprise: determining one or more actions to be taken by the communication device to minimize the expected total disconnection time given a maximum allowable time for the communication device to reach the goal position; and determining the trajectory basedon at least the one or more actions. In some embodiments, an action may be defined by at least one of: a target velocity at which the communication device is going to travel; a steering angle at which the communication device is going to travel; and a time for which the communication device is going to travel. In some embodiments, when the trajectory to be planned is a reactive trajectory, the one or more actions may bedetermined by using a Bayesian Reinforcement Learning (RL) algorithm based on one ormore states of the communication device and / or one or more samples drawn from the cellular world model. In some embodiments, the one or more actions may be determined by using asimulation-based Monte Carlo Tree Search (MCTS) algorithm based on one or morestates of the communication device. In some embodiments, at the start of each MCTS trial, the method 1300 may further comprise: sampling a single fixed state and model dynamics; and using the single fixed state and model dynamics throughout the MCTS trial. In some embodiments, the single fixed state and model dynamics may be sampled by: sampling, from the cellular world model, a cellular world instance that is to be used by the MCTS algorithm in determining how the one or more states of the communicationdevice should evolve; and initializing a new cellular transition system with a randominitial state. In some embodiments, the one or more states of the communication device may comprise at least one of: the time; the position of the communication device; the heading angle of the communication device; the cell that is associated with the communication device; and the RSRP measurements for one or more cells. In some embodiments, an action may be enabled only when taking the action leaves enough time for the communication device to reach the goal position without exceeding the maximum allowable time. In some embodiments, the step of planning the trajectory may further comprise: determining whether or not a candidate trajectory,which is output by the Bayesian RL algorithm or the simulation-based MCTS algorithm,meets one or more requirements. In some embodiments, the method 1300 may furthercomprise: in response to determining that no candidate trajectory meets the one ormore requirements, reporting at least one of: no trajectory can be planned provided the one or more requirements because of no enough knowledge of the cellular environment; and a level of knowledge of the cellular environment. In some embodiments, after the step of planning the trajectory, the method 1300 may further comprise at least one of: transmitting, to the communication device, the planned trajectory; reporting, to a traffic management system, the planned trajectory; and receiving, from the traffic management system, a message indicating whether theplanned trajectory is allowed or rejected. In some embodiments, the method 1300 mayfurther comprise: re-planning a trajectory in response to receiving the message indicating that the planned trajectory is rejected. In some embodiments, the method1300 may further comprise: receiving, from the communication device, a messageindicating new RSRP measurements; and updating one or more per-cell datasets based on the new RSRP measurements. Fig. 14 schematically shows an embodiment of an arrangement 1400 which maybe used in devices according to an embodiment of the present disclosure. Comprised inthe arrangement 1400 are a processing unit 1406, e.g., with a Digital Signal Processor(DSP) or a Central Processing Unit (CPU). The processing unit 1406 may be a single unitor a plurality of units to perform different actions of procedures described herein. The arrangement 1400 may also comprise an input unit 1402 for receiving signals from other entities, and an output unit 1404 for providing signal(s) to other entities. The input unit 1402 and the output unit 1404 may be arranged as an integrated entity or as separate entities. Furthermore, the arrangement 1400 may comprise at least one computer program product 1408 in the form of a non-volatile or volatile memory, e.g., an Electrically Erasable Programmable Read-Only Memory (EEPROM), a flash memory and / or a hard drive. The computer program product 1408 comprises a computer program 1410, which comprises code / computer readable instructions, which whenexecuted by the processing unit 1406 in the arrangement 1400 causes the arrangement1400 and / or the communication device in which it is comprised to perform the actions,e.g., of the procedure described earlier in conjunction with Fig. 4 through Fig. 13 or anyother variant. The computer program 1410 may be configured as a computer program codestructured in a computer program module 1410A. Hence, in an exemplifyingembodiment when the arrangement 1400 is used in a device for generating a cellularworld model, the code in the computer program of the arrangement 1400 includes: amodule 1410A configured to generate a cellular world model indicating a belief aboutwhat a true cellular environment is. Additionally or alternatively, the computer program 1410 may be further configured as a computer program code structured in a computer program module 1410B. Hence, in an exemplifying embodiment when the arrangement 1400 is used in a device for planning a trajectory for a communication device, the code in the computerprogram of the arrangement 1400 includes: a module 1410B configured to plan atrajectory for the communication device in a cellular environment based on at least a cellular world model that indicates a belief about what the cellular environment is. The computer program modules could essentially perform the actions of the flowillustrated in Fig. 4 through Fig. 13, to emulate the devices. In other words, when thedifferent computer program modules are executed in the processing unit 1406, they maycorrespond to different modules in the devices.Although the code means in the embodiments disclosed above in conjunction withFig. 14 are implemented as computer program modules which when executed in theprocessing unit causes the arrangement to perform the actions described above in conjunction with the figures mentioned above, at least one of the code means may in alternative embodiments be implemented at least partly as hardware circuits. The processor may be a single CPU (Central processing unit), but could also comprise two or more processing units. For example, the processor may include general purpose microprocessors; instruction set processors and / or related chips sets and / or special purpose microprocessors such as Application Specific Integrated Circuit (ASICs). The processor may also comprise board memory for caching purposes. The computer program may be carried by a computer program product connected to the processor. The computer program product may comprise a computer readable medium on which the computer program is stored. For example, the computer program product may be a flash memory, a Random-access memory (RAM), a Read-Only Memory (ROM), or an EEPROM, and the computer program modules described above could in alternative embodiments be distributed on different computer program products in the form of memories within the devices. Fig. 15 shows an example of a communication system QQ100 in accordance withsome embodiments. In the example, the communication system QQ100 includes a telecommunication network QQ102 that includes an access network QQ104, such as a radio access network(RAN), and a core network QQ106, which includes one or more core network nodesQQ108. The access network QQ104 includes one or more access network nodes, such asnetwork nodes QQ110A and QQ110B (one or more of which may be generally referred toas network nodes QQ110), or any other similar 3rdGeneration Partnership Project (3GPP) access nodes or non-3GPP access points. Moreover, as will be appreciated by those of skill in the art, a network node is not necessarily limited to an implementation in which a radio portion and a baseband portion are supplied and integrated by a single vendor. Thus, it will be understood that network nodes include disaggregated implementations or portions thereof. For example, in some embodiments, the telecommunication networkQQ102 includes one or more Open-RAN (ORAN) network nodes. An ORAN network nodeis a node in the telecommunication network QQ102 that supports an ORAN specification (e.g., a specification published by the O-RAN Alliance, or any similar organization) and may operate alone or together with other nodes to implement one or more functionalities of any node in the telecommunication network QQ102, including one or more network nodes QQ110 and / or core network nodes QQ108. Examples of an ORAN network node include an open radio unit (O-RU), an opendistributed unit (O-DU), an open central unit (O-CU), including an O-CU control plane (O-CU-CP) or an O-CU user plane (O-CU-UP), a RAN intelligent controller (near-real time or non-real time) hosting software or software plug-ins, such as a near-real time control application (e.g., xApp) or a non-real time control application (e.g., rApp), or any combination thereof (the adjective “open” designating support of an ORAN specification). The network node may support a specification by, for example, supporting an interface defined by the ORAN specification, such as an A1, F1, W1, E1, E2, X2, Xn interface, an open fronthaul user plane interface, or an open fronthaul management plane interface. Moreover, an ORAN access node may be a logical node in a physical node. Furthermore, an ORAN network node may be implemented in a virtualization environment (described further below) in which one or more network functions are virtualized. For example, the virtualization environment may include an O-Cloud computing platform orchestrated by a Service Management and Orchestration Framework via an O-2 interface defined by the O-RAN Alliance or comparable technologies. The network nodes QQ110 facilitate direct or indirect connection of user equipment (UE), such as by connecting UEs QQ112A,QQ112B, QQ112C, and QQ112D (one or more of which may be generally referred to asUEs QQ112) to the core network QQ106 over one or more wireless connections. Example wireless communications over a wireless connection include transmitting and / or receiving wireless signals using electromagnetic waves, radio waves, infrared waves, and / or other types of signals suitable for conveying information without the use of wires, cables, or other material conductors. Moreover, in different embodiments, the communication system QQ100 may include any number of wired or wireless networks, network nodes, UEs, and / or any other components or systems that may facilitate or participate in the communication of data and / or signals whether via wired or wireless connections. The communication system QQ100 may include and / or interface with any type of communication, telecommunication, data, cellular, radio network, and / or other similar type of system. The UEs QQ112 may be any of a wide variety of communication devices, including wireless devices arranged, configured, and / or operable to communicate wirelessly with the network nodes QQ110 and other communication devices. Similarly, the network nodes QQ110 are arranged, capable, configured, and / or operable to communicate directly or indirectly with the UEs QQ112 and / or with other network nodes or equipment in the telecommunication network QQ102 to enable and / or provide network access, such as wireless network access, and / or to perform other functions, such as administration in the telecommunication network QQ102. In the depicted example, the core network QQ106 connects the network nodesQQ110 to one or more host computing systems, such as host QQ116. These connectionsmay be direct or indirect via one or more intermediary networks or devices. In other examples, network nodes may be directly coupled to hosts. The core network QQ106 includes one more core network nodes (e.g., core network node QQ108) that are structured with hardware and software components. Features of these components may be substantially similar to those described with respect to the UEs, network nodes, and / or hosts, such that the descriptions thereof are generally applicable to the corresponding components of the core network node QQ108. Example core network nodes include functions of one or more of a Mobile Switching Center (MSC), Mobility Management Entity (MME), Home Subscriber Server (HSS), Access and Mobility Management Function (AMF), Session Management Function (SMF), Authentication Server Function (AUSF), Subscription Identifier De-concealing function (SIDF), Unified Data Management (UDM), Security Edge Protection Proxy (SEPP), Network Exposure Function (NEF), and / or a User Plane Function (UPF). The host QQ116 may be under the ownership or control of a service provider other than an operator or provider of the access network QQ104 and / or the telecommunication network QQ102. The host QQ116 may host a variety of applications to provide one or more service. Examples of such applications include live and pre- recorded audio / video content, data collection services such as retrieving and compiling data on various ambient conditions detected by a plurality of UEs, analytics functionality, social media, functions for controlling or otherwise interacting with remote devices, functions for an alarm and surveillance center, or any other such function performed by a server. As a whole, the communication system QQ100 of Fig. 15 enables connectivitybetween the UEs, network nodes, and hosts. In that sense, the communication system may be configured to operate according to predefined rules or procedures, such as specific standards that include, but are not limited to: Global System for Mobile Communications (GSM); Universal Mobile Telecommunications System (UMTS); Long Term Evolution (LTE), and / or other suitable 2G, 3G, 4G, 5G standards, or any applicable future generation standard (e.g., 6G); wireless local area network (WLAN) standards, such as the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standards (WiFi); and / or any other appropriate wireless communication standard, such as the Worldwide Interoperability for Microwave Access (WiMax), Bluetooth, Z-Wave, Near Field Communication (NFC) ZigBee, LiFi, and / or any low-power wide-area network (LPWAN) standards such as LoRa and Sigfox. In some examples, the telecommunication network QQ102 is a cellular network that implements 3GPP standardized features. Accordingly, the telecommunications network QQ102 may support network slicing to provide different logical networks to different devices that are connected to the telecommunication network QQ102. For example, the telecommunications network QQ102 may provide Ultra Reliable Low Latency Communication (URLLC) services to some UEs, while providing Enhanced Mobile Broadband (eMBB) services to other UEs, and / or Massive Machine Type Communication (mMTC) / Massive IoT services to yet further UEs. In some examples, the UEs QQ112 are configured to transmit and / or receive information without direct human interaction. For instance, a UE may be designed to transmit information to the access network QQ104 on a predetermined schedule, when triggered by an internal or external event, or in response to requests from the accessnetwork QQ104. Additionally, a UE may be configured for operating in single- or multi-RAT or multi-standard mode. For example, a UE may operate with any one or combination of Wi-Fi, NR (New Radio) and LTE, i.e. being configured for multi-radio dual connectivity (MR-DC), such as E-UTRAN (Evolved-UMTS Terrestrial Radio AccessNetwork) New Radio - Dual Connectivity (EN-DC).In the example, the hub QQ114 communicates with the access network QQ104 tofacilitate indirect communication between one or more UEs (e.g., UE QQ112C and / orQQ112D) and network nodes (e.g., network node QQ110B). In some examples, the hub QQ114 may be a controller, router, content source and analytics, or any of the other communication devices described herein regarding UEs. For example, the hub QQ114 may be a broadband router enabling access to the core network QQ106 for the UEs. As another example, the hub QQ114 may be a controller that sends commands or instructions to one or more actuators in the UEs. Commands or instructions may be received from the UEs, network nodes QQ110, or by executable code, script, process, or other instructions in the hub QQ114. As another example, the hub QQ114 may be a data collector that acts as temporary storage for UE data and, in some embodiments, may perform analysis or other processing of the data. As another example, the hub QQ114may be a content source. For example, for a UE that is a Virtual Reality (VR) device,display, loudspeaker, or other media delivery device, the hub QQ114 may retrieve VR assets, video, audio, or other media or data related to sensory information via a network node, which the hub QQ114 then provides to the UE either directly, after performing local processing, and / or after adding additional local content. In still another example, the hub QQ114 acts as a proxy server or orchestrator for the UEs, in particular if one or more of the UEs are low energy IoT devices. The hub QQ114 may have a constant / persistent or intermittent connection to the network node QQ110B. The hub QQ114 may also allow for a different communicationscheme and / or schedule between the hub QQ114 and UEs (e.g., UE QQ112C and / orQQ112D), and between the hub QQ114 and the core network QQ106. In other examples, the hub QQ114 is connected to the core network QQ106 and / or one or more UEs via awired connection. Moreover, the hub QQ114 may be configured to connect to a Machine-to-Machine (M2M) service provider over the access network QQ104 and / or to another UEover a direct connection. In some scenarios, UEs may establish a wireless connection with the network nodes QQ110 while still connected via the hub QQ114 via a wired or wireless connection. In some embodiments, the hub QQ114 may be a dedicated hub - that is, a hub whose primary function is to route communications to / from the UEs from / to the network node QQ110B. In other embodiments, the hub QQ114 may be anon-dedicated hub - that is, a device which is capable of operating to routecommunications between the UEs and network node QQ110B, but which is additionally capable of operating as a communication start and / or end point for certain data channels. Fig. 16 shows a UE QQ200 in accordance with some embodiments. The UEQQ200 presents additional details of some embodiments of the UE QQ112 of Fig. 15. Asused herein, a UE refers to a device capable, configured, arranged and / or operable to communicate wirelessly with network nodes and / or other UEs. Examples of a UE include,but are not limited to, a smart phone, mobile phone, cell phone, voice over IP (VoIP)phone, wireless local loop phone, desktop computer, personal digital assistant (PDA), wireless cameras, gaming console or device, music storage / playback device, wearable terminal device, wireless endpoint, mobile station, tablet, laptop, laptop-embedded equipment (LEE), laptop-mounted equipment (LME), an Augmented Reality (AR) or Virtual Reality (VR) device, wireless customer-premise equipment (CPE), vehicle, vehicle- mounted or vehicle embedded / integrated wireless device, etc. Other examples include any UE identified by the 3rdGeneration Partnership Project (3GPP), including a narrow band internet of things (NB-IoT) UE, a machine type communication (MTC) UE, and / or an enhanced MTC (eMTC) UE. A UE may support device-to-device (D2D) communication, for example by implementing a 3GPP standard for sidelink communication, Dedicated Short-Range Communication (DSRC), vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), or vehicle-to-everything (V2X). In other examples, a UE may not necessarily have a user in the sense of a human user who owns and / or operates the relevant device. Instead, a UE may represent a device that is intended for sale to, or operation by, a human user but which may not, or which may not initially, be associated with a specific human user (e.g., a smart sprinkler controller). Alternatively, a UE may represent a device that is not intended for sale to, or operation by, an end user but which may be associated with or operated for the benefit of a user (e.g., a smart power meter). The UE QQ200 includes processing circuitry QQ202 that is operatively coupled via a bus QQ204 to an input / output interface QQ206, a power source QQ208, a memory QQ210, a communication interface QQ212, and / or any other component, or any combination thereof. Certain UEs may utilize all or a subset of the components shown inFig. 16. The level of integration between the components may vary from one UE toanother UE. Further, certain UEs may contain multiple instances of a component, such as multiple processors, memories, transceivers, transmitters, receivers, etc. The processing circuitry QQ202 is configured to process instructions and data and may be configured to implement any sequential state machine operative to execute instructions stored as machine-readable computer programs in the memory QQ210. Theprocessing circuitry QQ202 may be implemented as one or more hardware-implementedstate machines (e.g., in discrete logic, field-programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), etc.); programmable logic together with appropriate firmware; one or more stored computer programs, general-purpose processors, such as a microprocessor or digital signal processor (DSP), together with appropriate software; or any combination of the above. For example, the processing circuitry QQ202 may include multiple central processing units (CPUs). In the example, the input / output interface QQ206 may be configured to provide an interface or interfaces to an input device, output device, or one or more input and / or output devices. Examples of an output device include a speaker, a sound card, a video card, a display, a monitor, a printer, an actuator, an emitter, a smartcard, another output device, or any combination thereof. An input device may allow a user to capture information into the UE QQ200. Examples of an input device include a touch-sensitive or presence-sensitive display, a camera (e.g., a digital camera, a digital video camera, a web camera, etc.), a microphone, a sensor, a mouse, a trackball, a directional pad, a trackpad, a scroll wheel, a smartcard, and the like. The presence-sensitive display may include a capacitive or resistive touch sensor to sense input from a user. A sensor may be, for instance, an accelerometer, a gyroscope, a tilt sensor, a force sensor, a magnetometer, an optical sensor, a proximity sensor, a biometric sensor, etc., or any combination thereof. An output device may use the same type of interface port as an input device. For example, a Universal Serial Bus (USB) port may be used to provide an input device and an output device. In some embodiments, the power source QQ208 is structured as a battery or battery pack. Other types of power sources, such as an external power source (e.g., an electricity outlet), photovoltaic device, or power cell, may be used. The power source QQ208 may further include power circuitry for delivering power from the power source QQ208 itself, and / or an external power source, to the various parts of the UE QQ200 via input circuitry or an interface such as an electrical power cable. Delivering power may be, for example, for charging of the power source QQ208. Power circuitry may perform any formatting, converting, or other modification to the power from the power source QQ208 to make the power suitable for the respective components of the UE QQ200 to which power is supplied. The memory QQ210 may be or be configured to include memory such as random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic disks, optical disks, hard disks, removable cartridges, flash drives, and so forth. In one example, the memory QQ210 includes one or more application programs QQ214, such as an operating system, web browser application, a widget, gadget engine, or other application, and corresponding data QQ216. The memory QQ210 may store, for use by the UE QQ200, any of a variety of various operating systems or combinations of operating systems. The memory QQ210 may be configured to include a number of physical drive units, such as redundant array of independent disks (RAID), flash memory, USB flash drive, external hard disk drive, thumb drive, pen drive, key drive, high-density digital versatile disc (HD-DVD) optical disc drive, internal hard disk drive, Blu-Ray optical disc drive, holographic digital data storage (HDDS) optical disc drive, external mini-dual in- line memory module (DIMM), synchronous dynamic random access memory (SDRAM), external micro-DIMM SDRAM, smartcard memory such as tamper resistant module in the form of a universal integrated circuit card (UICC) including one or more subscriber identity modules (SIMs), such as a USIM and / or ISIM, other memory, or any combination thereof. The UICC may for example be an embedded UICC (eUICC), integrated UICC (iUICC) or a removable UICC commonly known as ‘SIM card.’ The memory QQ210 may allow the UE QQ200 to access instructions, application programs and the like, stored on transitory or non-transitory memory media, to off-load data, or to upload data. An article of manufacture, such as one utilizing a communication system may be tangibly embodied as or in the memory QQ210, which may be or comprise a device-readable storage medium. The processing circuitry QQ202 may be configured to communicate with an access network or other network using the communication interface QQ212. The communication interface QQ212 may comprise one or more communication subsystemsand may include or be communicatively coupled to an antenna QQ222. Thecommunication interface QQ212 may include one or more transceivers used to communicate, such as by communicating with one or more remote transceivers of another device capable of wireless communication (e.g., another UE or a network node in an access network). Each transceiver may include a transmitter QQ218 and / or a receiver QQ220 appropriate to provide network communications (e.g., optical, electrical, frequency allocations, and so forth). Moreover, the transmitter QQ218 and receiver QQ220 may be coupled to one or more antennas (e.g., antenna QQ222) and may share circuit components, software or firmware, or alternatively be implemented separately. In the illustrated embodiment, communication functions of the communication interface QQ212 may include cellular communication, Wi-Fi communication, LPWAN communication, data communication, voice communication, multimedia communication, short-range communications such as Bluetooth, near-field communication, location- based communication such as the use of the global positioning system (GPS) to determine a location, another like communication function, or any combination thereof. Communications may be implemented in according to one or more communication protocols and / or standards, such as IEEE 802.11, Code Division Multiplexing Access (CDMA), Wideband Code Division Multiple Access (WCDMA), GSM, LTE, New Radio (NR), UMTS, WiMax, Ethernet, transmission control protocol / internet protocol (TCP / IP), synchronous optical networking (SONET), Asynchronous Transfer Mode (ATM), QUIC, Hypertext Transfer Protocol (HTTP), and so forth. Regardless of the type of sensor, a UE may provide an output of data captured by its sensors, through its communication interface QQ212, via a wireless connection to a network node. Data captured by sensors of a UE can be communicated through a wireless connection to a network node via another UE. The output may be periodic (e.g., once every 15 minutes if it reports the sensed temperature), random (e.g., to even out the load from reporting from several sensors), in response to a triggering event (e.g., when moisture is detected an alert is sent), in response to a request (e.g., a user initiated request), or a continuous stream (e.g., a live video feed of a patient). As another example, a UE comprises an actuator, a motor, or a switch, related to a communication interface configured to receive wireless input from a network node via a wireless connection. In response to the received wireless input the states of the actuator, the motor, or the switch may change. For example, the UE may comprise a motor that adjusts the control surfaces or rotors of a drone in flight according to the received input or to a robotic arm performing a medical procedure according to the received input. A UE, when in the form of an Internet of Things (IoT) device, may be a device for use in one or more application domains, these domains comprising, but not limited to, city wearable technology, extended industrial application and healthcare. Non-limiting examples of such an IoT device are a device which is or which is embedded in: a connected refrigerator or freezer, a TV, a connected lighting device, an electricity meter, a robot vacuum cleaner, a voice controlled smart speaker, a home security camera, a motion detector, a thermostat, a smoke detector, a door / window sensor, a flood / moisture sensor, an electrical door lock, a connected doorbell, an air conditioning system like a heat pump, an autonomous vehicle, a surveillance system, a weather monitoring device, a vehicle parking monitoring device, an electric vehicle charging station, a smart watch, a fitness tracker, a wearable for tactile augmentation or sensoryenhancement, a water sprinkler, an animal- or item-tracking device, a sensor formonitoring a plant or animal, an industrial robot, an Unmanned Aerial Vehicle (UAV), and any kind of medical device, like a heart rate monitor or a remote controlled surgical robot. A UE in the form of an IoT device comprises circuitry and / or software in dependence of the intended application of the IoT device in addition to other components as described in relation to the UE QQ200 shown in Fig. 16. As yet another specific example, in an IoT scenario, a UE may represent a machine or other device that performs monitoring and / or measurements, and transmits the results of such monitoring and / or measurements to another UE and / or a network node. The UE may in this case be an M2M device, which may in a 3GPP context be referred to as an MTC device. As one particular example, the UE may implement the 3GPP NB-IoT standard. In other scenarios, a UE may represent a vehicle, such as a car, a bus, a truck, a ship and an airplane, or other equipment that is capable of monitoringand / or reporting on its operational status or other functions associated with its operation.In practice, any number of UEs may be used together with respect to a single use case. For example, a first UE might be or be integrated in a drone and provide the drone’s speed information (obtained through a speed sensor) to a second UE that is a remote controller operating the drone. When the user makes changes from the remote controller, the first UE may adjust the throttle on the drone (e.g. by controlling an actuator) to increase or decrease the drone’s speed. The first and / or the second UE can also include more than one of the functionalities described above. For example, a UE might comprise the sensor and the actuator, and handle communication of data for both the speed sensor and the actuators. Fig. 17 shows a network node QQ300 in accordance with some embodiments. Asused herein, network node refers to equipment capable, configured, arranged and / or operable to communicate directly or indirectly with a UE and / or with other network nodes or equipment, in a telecommunication network. Examples of network nodes include, but are not limited to, access points (APs) (e.g., radio access points), base stations (BSs) (e.g., radio base stations, Node Bs, evolved Node Bs (eNBs) and NR NodeBs (gNBs)), O-RAN nodes or components of an O-RAN node (e.g., O-RU, O-DU, O- CU). Base stations may be categorized based on the amount of coverage they provide (or, stated differently, their transmit power level) and so, depending on the provided amount of coverage, may be referred to as femto base stations, pico base stations, micro base stations, or macro base stations. A base station may be a relay node or a relay donor node controlling a relay. A network node may also include one or more (or all) parts of a distributed radio base station such as centralized digital units, distributed units (e.g., in an O-RAN access node) and / or remote radio units (RRUs), sometimes referred to as Remote Radio Heads (RRHs). Such remote radio units may or may not be integrated with an antenna as an antenna integrated radio. Parts of a distributed radio base station may also be referred to as nodes in a distributed antenna system (DAS). Other examples of network nodes include multiple transmission point (multi-TRP) 5G access nodes, multi-standard radio (MSR) equipment such as MSR BSs, network controllers such as radio network controllers (RNCs) or base station controllers (BSCs), base transceiver stations (BTSs), transmission points, transmission nodes, multi- cell / multicast coordination entities (MCEs), Operation and Maintenance (O&M) nodes, Operations Support System (OSS) nodes, Self-Organizing Network (SON) nodes, positioning nodes (e.g., Evolved Serving Mobile Location Centers (E-SMLCs)), and / or Minimization of Drive Tests (MDTs). The network node QQ300 includes a processing circuitry QQ302, a memory QQ304, a communication interface QQ306, and a power source QQ308. The network node QQ300 may be composed of multiple physically separate components (e.g., a NodeB component and a RNC component, or a BTS component and a BSC component, etc.), which may each have their own respective components. In certain scenarios in which the network node QQ300 comprises multiple separate components (e.g., BTS and BSC components), one or more of the separate components may be shared amongseveral network nodes. For example, a single RNC may control multiple NodeBs. In sucha scenario, each unique NodeB and RNC pair, may in some instances be considered a single separate network node. In some embodiments, the network node QQ300 may be configured to support multiple radio access technologies (RATs). In such embodiments, some components may be duplicated (e.g., separate memory QQ304 for different RATs) and some components may be reused (e.g., a same antenna QQ310 may be shared by different RATs). The network node QQ300 may also include multiple sets of the various illustrated components for different wireless technologies integrated into network node QQ300, for example GSM, WCDMA, LTE, NR, WiFi, Zigbee, Z-wave, LoRaWAN, Radio Frequency Identification (RFID) or Bluetooth wireless technologies. These wireless technologies may be integrated into the same or different chip or set of chips and other components within network node QQ300. The processing circuitry QQ302 may comprise a combination of one or more of a microprocessor, controller, microcontroller, central processing unit, digital signal processor, application-specific integrated circuit, field programmable gate array, or any other suitable computing device, resource, or combination of hardware, software and / or encoded logic operable to provide, either alone or in conjunction with other network node QQ300 components, such as the memory QQ304, to provide network node QQ300 functionality. In some embodiments, the processing circuitry QQ302 includes a system on achip (SOC). In some embodiments, the processing circuitry QQ302 includes one or more of radio frequency (RF) transceiver circuitry QQ312 and baseband processing circuitry QQ314. In some embodiments, the radio frequency (RF) transceiver circuitry QQ312 and the baseband processing circuitry QQ314 may be on separate chips (or sets of chips), boards, or units, such as radio units and digital units. In alternative embodiments, part or all of RF transceiver circuitry QQ312 and baseband processing circuitry QQ314 may be on the same chip or set of chips, boards, or units. The memory QQ304 may comprise any form of volatile or non-volatile computer- readable memory including, without limitation, persistent storage, solid-state memory, remotely mounted memory, magnetic media, optical media, random access memory (RAM), read-only memory (ROM), mass storage media (for example, a hard disk), removable storage media (for example, a flash drive, a Compact Disk (CD) or a Digital Video Disk (DVD)), and / or any other volatile or non-volatile, non-transitory device- readable and / or computer-executable memory devices that store information, data, and / or instructions that may be used by the processing circuitry QQ302. The memory QQ304 may store any suitable instructions, data, or information, including a computer program, software, an application including one or more of logic, rules, code, tables, and / or other instructions capable of being executed by the processing circuitry QQ302 and utilized by the network node QQ300. The memory QQ304 may be used to store any calculations made by the processing circuitry QQ302 and / or any data received via the communication interface QQ306. In some embodiments, the processing circuitry QQ302 and memory QQ304 is integrated. The communication interface QQ306 is used in wired or wireless communication of signaling and / or data between a network node, access network, and / or UE. As illustrated, the communication interface QQ306 comprises port(s) / terminal(s) QQ316 to send and receive data, for example to and from a network over a wired connection. The communication interface QQ306 also includes radio front-end circuitry QQ318 that may be coupled to, or in certain embodiments a part of, the antenna QQ310. Radio front-end circuitry QQ318 comprises filters QQ320 and amplifiers QQ322. The radio front-end circuitry QQ318 may be connected to an antenna QQ310 and processing circuitry QQ302. The radio front-end circuitry may be configured to condition signals communicated between antenna QQ310 and processing circuitry QQ302. The radio front-end circuitry QQ318 may receive digital data that is to be sent out to other network nodes or UEs via a wireless connection. The radio front-end circuitry QQ318 may convert the digital data into a radio signal having the appropriate channel and bandwidth parameters using a combination of filters QQ320 and / or amplifiers QQ322. The radio signal may then be transmitted via the antenna QQ310. Similarly, when receiving data, the antenna QQ310 may collect radio signals which are then converted into digital data by the radio front- end circuitry QQ318. The digital data may be passed to the processing circuitry QQ302. In other embodiments, the communication interface may comprise different components and / or different combinations of components. In certain alternative embodiments, the network node QQ300 does not include separate radio front-end circuitry QQ318, instead, the processing circuitry QQ302 includes radio front-end circuitry and is connected to the antenna QQ310. Similarly, in some embodiments, all or some of the RF transceiver circuitry QQ312 is part of the communication interface QQ306. In still other embodiments, the communication interface QQ306 includes one or more ports or terminals QQ316, the radio front-end circuitry QQ318, and the RF transceiver circuitry QQ312, as part of a radio unit (not shown), and the communication interface QQ306 communicates with the baseband processing circuitry QQ314, which is part of a digital unit (not shown). The antenna QQ310 may include one or more antennas, or antenna arrays, configured to send and / or receive wireless signals. The antenna QQ310 may be coupled to the radio front-end circuitry QQ318 and may be any type of antenna capable of transmitting and receiving data and / or signals wirelessly. In certain embodiments, the antenna QQ310 is separate from the network node QQ300 and connectable to the network node QQ300 through an interface or port. The antenna QQ310, communication interface QQ306, and / or the processing circuitry QQ302 may be configured to perform any receiving operations and / or certain obtaining operations described herein as being performed by the network node. Any information, data and / or signals may be received from a UE, another network node and / or any other network equipment. Similarly, the antenna QQ310, the communication interface QQ306, and / or the processing circuitry QQ302 may be configured to perform any transmitting operations described herein as being performed by the network node. Any information, data and / or signals may be transmitted to a UE, another network node and / or any other network equipment. The power source QQ308 provides power to the various components of network node QQ300 in a form suitable for the respective components (e.g., at a voltage and current level needed for each respective component). The power source QQ308 may further comprise, or be coupled to, power management circuitry to supply the components of the network node QQ300 with power for performing the functionality described herein. For example, the network node QQ300 may be connectable to an external power source (e.g., the power grid, an electricity outlet) via an input circuitry or interface such as an electrical cable, whereby the external power source supplies power to power circuitry of the power source QQ308. As a further example, the power source QQ308 may comprise a source of power in the form of a battery or battery pack which is connected to, or integrated in, power circuitry. The battery may provide backup powershould the external power source fail.Embodiments of the network node QQ300 may include additional componentsbeyond those shown in Fig. 17 for providing certain aspects of the network node’sfunctionality, including any of the functionality described herein and / or any functionality necessary to support the subject matter described herein. For example, the network node QQ300 may include user interface equipment to allow input of information into the network node QQ300 and to allow output of information from the network node QQ300. This may allow a user to perform diagnostic, maintenance, repair, and other administrative functions for the network node QQ300. In some embodiments providing a core network node, such as core network node 108 of Fig. 15, some components, such as the radio front-end circuitry QQ318 and the RF transceiver circuitry QQ312 may be omitted. Fig. 18 is a block diagram illustrating a virtualization environment QQ400 in whichfunctions implemented by some embodiments may be virtualized. In the present context, virtualizing means creating virtual versions of apparatuses or devices which may include virtualizing hardware platforms, storage devices and networking resources. As used herein, virtualization can be applied to any device described herein, or components thereof, and relates to an implementation in which at least a portion of the functionality is implemented as one or more virtual components. Some or all of the functions described herein may be implemented as virtual components executed by one or morevirtual machines (VMs) implemented in one or more virtual environments QQ400 hostedby one or more of hardware nodes, such as a hardware computing device that operates as a network node, UE, core network node, or host. Further, in embodiments in whichthe virtual node does not require radio connectivity (e.g., a core network node or host),then the node may be entirely virtualized. In some embodiments, the virtualization environment QQ400 includes components defined by the O-RAN Alliance, such as an O- Cloud environment orchestrated by a Service Management and Orchestration Framework via an O-2 interface. Virtualization may facilitate distributed implementations of a network node, UE, core network node, or host. Applications QQ402 (which may alternatively be called software instances, virtual appliances, network functions, virtual nodes, virtual network functions, etc.) are run in the virtualization environment Q400 to implement some of the features, functions, and / or benefits of some of the embodiments disclosed herein. Hardware QQ404 includes processing circuitry, memory that stores software and / or instructions executable by hardware processing circuitry, and / or other hardware devices as described herein, such as a network interface, input / output interface, and so forth. Software may be executed by the processing circuitry to instantiate one or more virtualization layers QQ406 (also referred to as hypervisors or virtual machine monitors (VMMs)), provide VMs QQ408a and QQ408b (one or more of which may be generally referred to as VMs QQ408), and / or perform any of the functions, features and / or benefits described in relation with some embodiments described herein. The virtualization layer QQ406 may present a virtual operating platform that appears like networking hardware to the VMs QQ408. The VMs QQ408 comprise virtual processing, virtual memory, virtual networking or interface and virtual storage, and may be run by a corresponding virtualization layer QQ406. Different embodiments of the instance of a virtual appliance QQ402 may be implemented on one or more of VMs QQ408, and the implementations may be made in different ways. Virtualization of the hardware is in some contexts referred to as network function virtualization (NFV). NFV may be used to consolidate many network equipment types onto industry standard high volume server hardware, physical switches, and physical storage, which can be located in data centers, and customer premise equipment. In the context of NFV, a VM QQ408 may be a software implementation of a physical machine that runs programs as if they were executing on a physical, non- virtualized machine. Each of the VMs QQ408, and that part of hardware QQ404 that executes that VM, be it hardware dedicated to that VM and / or hardware shared by thatVM with others of the VMs, forms separate virtual network elements. Still in the contextof NFV, a virtual network function is responsible for handling specific network functions that run in one or more VMs QQ408 on top of the hardware QQ404 and corresponds to the application QQ402. Hardware QQ404 may be implemented in a standalone network node with generic or specific components. Hardware QQ404 may implement some functions via virtualization. Alternatively, hardware QQ404 may be part of a larger cluster of hardware (e.g. such as in a data center or CPE) where many hardware nodes work together and are managed via management and orchestration QQ410, which, among others, oversees lifecycle management of applications QQ402. In some embodiments, hardware QQ404 is coupled to one or more radio units that each include one or more transmitters and one or more receivers that may be coupled to one or more antennas. Radio units may communicate directly with other hardware nodes via one or more appropriate network interfaces and may be used in combination with the virtual components to provide a virtual node with radio capabilities, such as a radio access node or a base station. In some embodiments, some signaling can be provided with the use of a control systemQQ412 which may alternatively be used for communication between hardware nodesand radio units. Although the computing devices described herein (e.g., UEs, network nodes) may include the illustrated combination of hardware components, other embodiments may comprise computing devices with different combinations of components. It is to be understood that these computing devices may comprise any suitable combination of hardware and / or software needed to perform the tasks, features, functions and methods disclosed herein. Determining, calculating, obtaining or similar operations described herein may be performed by processing circuitry, which may process information by, for example, converting the obtained information into other information, comparing the obtained information or converted information to information stored in the network node, and / or performing one or more operations based on the obtained information or converted information, and as a result of said processing making a determination. Moreover, while components are depicted as single boxes located within a larger box, or nested within multiple boxes, in practice, computing devices may comprise multiple different physical components that make up a single illustrated component, and functionality may be partitioned between separate components. For example, a communication interface may be configured to include any of the components described herein, and / or the functionality of the components may be partitioned between the processing circuitry and the communication interface. In another example, non- computationally intensive functions of any of such components may be implemented in software or firmware and computationally intensive functions may be implemented in hardware. In certain embodiments, some or all of the functionality described herein may be provided by processing circuitry executing instructions stored on in memory, which in certain embodiments may be a computer program product in the form of a non- transitory computer-readable storage medium. In alternative embodiments, some or all of the functionality may be provided by the processing circuitry without executing instructions stored on a separate or discrete device-readable storage medium, such as in a hard-wired manner. In any of those particular embodiments, whether executing instructions stored on a non-transitory computer-readable storage medium or not, the processing circuitry can be configured to perform the described functionality. The benefits provided by such functionality are not limited to the processing circuitry alone or to other components of the computing device, but are enjoyed by the computing device as a whole, and / or by end users and a wireless network generally. The present disclosure is described above with reference to the embodiments thereof. However, those embodiments are provided just for illustrative purpose, rather than limiting the present disclosure. The scope of the disclosure is defined by the attached claims as well as equivalents thereof. Those skilled in the art can make various alternations and modifications without departing from the scope of the disclosure, which all fall into the scope of the disclosure.

Claims

Claims1. A method (1200) for generating a cellular world model, the method (1200)comprising: generating (S520, S1210) a cellular world model indicating a belief aboutwhat a true cellular environment is.

2. The method (1200) of claim 1, wherein the cellular world model is able to be usedfor planning a trajectory for a communication device (100) in the true cellularenvironment.

3. The method (1200) of claim 1 or 2, wherein the cellular world model comprises atleast one of: -one or more per-cell link gain models;- one or more sampled per-cell link gain models;- one or more per-cell Signal-to-Interference-plus-Noise Ratio (SINR) models;- one or more sampled per-cell SINR models; and- one or more variables.

4. The method (1200) of claim 3, wherein the one or more per-cell link gain models,the one or more per-cell SINR models, and their sampled versions are spatial distribution models.

5. The method (1200) of claim 3 or 4, wherein the one or more per-cell link gainmodels and / or the one or more per-cell SINR models are probabilistic models that are fit to one or more per-cell datasets (^^) associated with corresponding cells, respectively.

6. The method (1200) of claim 5, wherein a probabilistic model is fit to a per-celldataset (^^) by: optimizing one or more hyper parameters of the probabilistic model by using a gradient descent algorithm on the log marginal likelihood.

7. The method (1200) of claim 5 or 6, wherein a per-cell dataset (^^) associatedwith a cell comprises one or more entries, each of which comprises at least one of: -an identifier (ID) of the cell;- a Reference Signal Received Power (RSRP) measurement for the cell;- an SINR measurement for the cell;- a position where the RSRP measurement and / or the SINR measurement is taken;and -a timestamp when the RSRP measurement and / or the SINR measurement istaken.

8. The method (1200) of any of claims 5 to 7, wherein an entry in a per-cell dataset(^^) is removed from the per-cell dataset (^^) when the entry has a measurement that differs from another measurement by a specified margin, wherein the measurement is taken at a first position, wherein the other measurement is taken at the first position or a second position that is separated from the first position by a specified distance or shorter, but later than the measurement.

9. The method (1200) of any of claims 5 to 8, wherein at least one entry in a per-cell dataset (^^) has a measurement that is a weighted average of at least one of: -measurements that are taken at different times; and- measurements that are taken at different positions that are separated from eachother by a specified distance or shorter.

10. The method (1200) of any of claims 5 to 9, wherein before a probabilistic model isfit to a per-cell dataset (^^), one or more entries in the per-cell dataset (^^) areconverted by: determining the position of an antenna associated with the corresponding cell; and converting one or more coordinates of one or more positions in the one or more entries into coordinates in a coordinate system with its origin being the position of the antenna.

11. The method (1200) of any of claims 5 to 10, wherein the probabilistic model is aGaussian process (GP) model having a kernel function that is either a single kernel function or a combination of multiple kernel functions.

12. The method (1200) of claim 11, wherein the multiple kernel functions comprise atleast a first kernel function based on a polar coordinate system and a second kernelfunction based on a Cartesian coordinate system.

13. The method (1200) of claim 12, wherein the first kernel function is determined by:^POLAR(^, ^^) = ^^ ^α ^^ ^angular(θ, θ^) + α ^^ ^^^^^^^ (ρ, ρ^)^, where ^POLAR(∙,∙) is the first kernel function, ^angular(∙,∙) is a GP kernel function thatapplies in the is a GPfunction that applies in the radial axis of the polar coordinate system, θ and θ^ areangular coordinates of positions ^ and ^ ^ in the polar coordinate system, respectively, ρand ρ^ are radial coordinates of ^ and ^^ in the polar coordinate system, respectively,and ^^, α^^, α^^, are weighting factors of the first kernel function.

14. The method (1200) of claim 13, wherein ^^^^^^^(∙,∙) is a Radial Basis Function(RBF).

15. The method (1200) of claim 13 or 14, wherein ^angular(∙,∙) is one of:^^ ^^- ^angular(θ, θ^) = ^^ ^2 sin^^, where ^^is the 2^-Wendland function, andsin(∙) is the sine function;− ^^)^ , where ^^is the ^^-Wendlandacos(∙) is the arccosine function, and cos(∙) is the cosine function; and- an RBF.

16. The method (1200) of any of claims 12 to 15, wherein the second kernel functionhas a length-scale parameter with a value less than that of the first kernel function.

17. The method (1200) of any of claims 12 to 16, wherein the second kernel functionis a Matérn kernel function.

18. The method (1200) of any of claims 3 to 17, wherein at least one of the multipleper-cell SINR models is determined by: γ ( ^(x) = ^ ^ρ^x) ∑ ^ ^^ ^ ^ ^^^^ ∙ ρ^^(x)^where γ (∙) is the ^ is a position at whichSINR for the cell ^ is to bewhere ρ^(∙) is the per-cell link gain model for the cell ^, ρ^^(∙) is the per-cell linkgain model for the cell c′ that is different from the cell ^, ^^^ is a binary variable subjectto a Bernoulli distribution with a probability of ^^^ for its value “1” and a probability of1 − ^^^ for its value “0”, and ^^^ is the utilization factor for the cell c′,where ∑^^ ^ ^ (∙) is a summation function for all cells c′ that are not the cell ^,where ^[∙] is a function for calculating the mathematical expectation of itsoperand.

19. The method (1200) of claims 3 to 18, wherein the one or more variables compriseat least one of: -a utilization factor for at least one cell; and- one or more variables associated with a handover procedure.

20. The method (1200) of claim 19, wherein the one or more variables associatedwith a handover procedure comprise at least one of:- a handover time-to-trigger;- a handover hysteresis;- a successful handover interruption time;- a failed handover interruption time;- one or more parameters for a probability distribution of a successful handoverinterruption time; and -one or more parameters for a probability distribution of a failed handoverinterruption time.

21. The method (1200) of claim 19 or 20, wherein the one or more variables compriseat least one of: -a variable that has a known true value;- a variable that has an assumed value;- a variable that has a probability distribution over possible values; and- a variable that has a worst-case value.

22. The method (1200) of any of claims 3 to 21, wherein a sampled per-cell link gainmodel associated with a cell is an instance of a possible true underlying link gain spatial distribution.

23. The method (1200) of claim 22, wherein the instance is drawn from acorresponding per-cell link gain model associated with the same cell.

24. The method (1200) of claim 22 or 23, wherein the instance is drawn at a set ofone or more position points.

25. The method (1200) of claim 24, wherein the set of position points cover anoperating area in which a trajectory is to be planned.

26. The method (1200) of any of claims 23 to 25, wherein a first number of sampledper-cell link gain models associated with a cell are stored and / or used in place of thecorresponding per-cell link gain model associated with the same cell.

27. The method (1200) of any of claims 3 to 26, wherein a sampled per-cell SINRmodel is determined by: ^^(x) = ^^^(x)∑^^ ^ ^ ^^^^ ∙ ^^^(x)^where ^^(∙) is the the cell ^, x is a position atwhich the SINR for the cell ^ is to be determined,where ^^^ (∙) is a sampled per-cell link gain model for the cell ^, ^^^^(∙) is a sampledper-cell link gain model for the cell c′ that is different from the cell ^, and ^^^ is theutilization factor for the cell c′, where ∑^^ ^ ^ (∙) is a summation function for all cells c′ that are different from thecell ^.

28. (1200) of any of claims 5 to 27, wherein the one or more per-celldatasets (^^) are continually updated over time as more measurements are obtained.

29. The method (1200) of any of claims 2 to 28, wherein the communication device(100) is an Unmanned Aerial Vehicle (UAV) (100-1).

30. A device (145, 1400) for generating a cellular world model, the device (145, 1400)comprising: aprocessor (1406);a memory (1408) storing instructions which, when executed by the processor(1406), cause the device (145, 1400) to:generate a cellular world model indicating a belief about what a truecellular environment is.

31. The device (145, 1400) of claim 30, wherein the instructions, when executed bythe processor (1406), further cause the device (145, 1400) to perform the method (1200)of any of claims 2 to 29.

32. A method (1300) for planning a trajectory for a communication device (100), themethod (1300) comprising:planning (S610, S1310) a trajectory for the communication device (100) in acellular environment based on at least a cellular world model that indicates a belief aboutwhat the cellular environment is.

33. The method (1300) of claim 32, wherein the cellular world model is generatedaccording to the method (1200) of any of claims 1 to 29.

34. The method (1300) of claim 32 or 33, wherein the trajectory is one of:- a fixed trajectory that cannot be changed during its execution; and- a reactive trajectory that can be changed during its execution.

35. The method (1300) of any of claims 32 to 34, wherein before and / or during thestep of planning (S610, S1310) the trajectory, the method (1300) further comprises:drawing, for each cell, a specified number of samples from a corresponding per- cell link gain model.

36. The method (1300) of claim 35, wherein before and / or during the step ofplanning (S610, S1310) the trajectory, the method (1300) further comprises:determining a sample index for each cell randomly; selecting the sample at the sample index from the specified number of samplesdrawn for each cell.

37. The method (1300) of any of claims 32 to 36, wherein before and / or during thestep of planning (S610, S1310) the trajectory, the method (1300) further comprises:determining a sampled per-cell SINR model by: ^ ( ) ^(x) =^^^x ∑^^ ^ ^ ^^^^ ∙ ^^^(x)^where ^^(∙) is the the cell ^, x is a position atwhich the SINR for the cell ^where ^^^ (∙) is the sampled per-cell link gain model for the cell ^, ^^^(∙) is thesampled per-cell link gain model for the cell c′ that is different from the cell ^, and ^^^ isthe utilization factor for the cell c′, where ∑^^ ^ ^ (∙) is a summation function for all cells c′ that are different from thecell ^.

38. The method (1300) of any of claims 32 to 37, wherein before and / or during thestep of planning (S610, S1310) the trajectory, the method (1300) further comprises:drawing a sample from a variable in the cellular world model according to its probability distribution when the variable is a random variable.

39. The method (1300) of any of claims 32 to 38, wherein before the step of planning(S610, S1310) the trajectory, the method (1300) further comprises obtaining at leastone of: -an operating area and / or an operating space in which the trajectory for thecommunication device (100) is to be planned; -one or more indicators, each of which indicates an area and / or space that thecommunication device (100) is not allowed to enter; and -one or more requirements on the trajectory to be planned.

40. The method (1300) of claim 39, wherein the operating area is a two-dimensional(2D) polygon bounding an area in which the communication device (100) is allowed tooperate.

41. The method (1300) of claim 39, wherein the operating space is a three-dimensional (3D) polyhedron bounding a space in which the communication device (100)is allowed to operate.

42. The method (1300) of any of claims 39 to 41, wherein the one or morerequirements comprise at least one of: -a desired operating altitude for the communication device (100);- a start position;- a goal position;- one or more positions that the communication device (100) is to visit at specifiedtimes; -a minimum SINR for the communication device (100) to maintain its connectionto the network; -a maximum allowable disconnection time;- a maximum allowable ratio of the total disconnection time to the total time fortrajectory execution; -a minimum probability that the total disconnection time never exceeds amaximum disconnection time; -a maximum allowable cost that is calculated based on a combination of the totaltime for trajectory execution and the total disconnection time; and- a maximum allowable time for the communication device (100) to reach the goalposition.

43. The method (1300) of any of claims 32 to 42, wherein before and / or during thestep of planning the trajectory, the method (1300) further comprises:initializing a cellular transition system that is able to track a cellular state of thecommunication device (100) along a trajectory.

44. The method (1300) of claim 43, wherein the cellular transition system is initializedwith a cellular world instance sampled from the cellular world model.

45. The method (1300) of claim 43 or 44, wherein the cellular transition system isable to determine the cellular state of the communication device (100) along the trajectory based on one or more inputs comprising at least one of: -one or more indicators, indicating a series of time when the communicationdevice (100) is travelling along the trajectory; -one or more indicators, each of which indicates an SINR vector comprising oneor more SINR values for one or more cells sampled at a position along the trajectory;- one or more indicators, each of which indicates a link gain vector comprising oneor more link gain values for one or more cells sampled at a position along the trajectory; and -one or more cellular world parameters.

46. The method (1300) of claim 45, wherein the one or more SINR values comprisedin the SINR vector are obtained by querying at least one of: -one or more corresponding per-cell SINR models; and- one or more corresponding sampled per-cell SINR models.

47. The method (1300) of claim 45 or 46, wherein the one or more link gain valuescomprised in the link gain vector are obtained by querying at least one of: -one or more corresponding per-cell link gain models; and- one or more corresponding sampled per-cell link gain models.

48. The method (1300) of any of claims 43 to 47, wherein the cellular state of thecommunication device (100) along the trajectory comprises at least one of: -one or more indicators, each of which indicates a cell, with which thecommunication device (100) is associated at a specified time; and -one or more indicators, each of which indicates a disconnection time during aspecified period.

49. The method (1300) of any of claims 43 to 48, wherein the cellular state isdetermined based on at least one of one or more handovers and one or more radio link failures (RLFs), wherein a handover from a first cell to a second cell for the communication device (100) is determined to occur when the second cell has a higher link gain than that of the first cell by more than a threshold value, which is indicated by a handover hysteresis variable, for at least a time period, which is indicated by a handover time-to-trigger variable; and wherein an RLF is determined to occur when an SINR of a currently associated cell falls below a minimum level.

50. The method (1300) of any of claims 32 to 49, wherein the step of planning (S610,S1310) the trajectory comprises: determining one or more actions to be taken by the communication device (100) to minimize the expected total disconnection time given a maximum allowable time for the communication device (100) to reach the goal position; and determining the trajectory based on at least the one or more actions.

51. The method (1300) of claim 50, wherein an action is defined by at least one of:- a target velocity at which the communication device (100) is going to travel;- a steering angle at which the communication device (100) is going to travel; and- a time for which the communication device (100) is going to travel.

52. The method (1300) of claim 50 or 51, wherein when the trajectory to be plannedis a reactive trajectory, the one or more actions are determined by using a BayesianReinforcement Learning (RL) algorithm based on one or more states of thecommunication device (100) and / or one or more samples drawn from the cellular world model.

53. The method (1300) of any of claims 50 to 52, wherein the one or more actionsare determined by using a simulation-based Monte Carlo Tree Search (MCTS) algorithmbased on one or more states of the communication device (100).

54. The method (1300) of claim 53, wherein at the start of each MCTS trial, themethod (1300) further comprises:sampling a single fixed state and model dynamics; and using the single fixed state and model dynamics throughout the MCTS trial.

55. The method (1300) of claim 54, wherein the single fixed state and modeldynamics is sampled by: sampling, from the cellular world model, a cellular world instance that is to be used by the MCTS algorithm in determining how the one or more states of the communication device (100) should evolve; and initializing a new cellular transition system with a random initial state.

56. The method (1300) of claim 54 or 55, wherein the one or more states of thecommunication device (100) comprise at least one of: -the time;- the position of the communication device (100);- the heading angle of the communication device (100);- the cell that is associated with the communication device (100); and- the RSRP measurements for one or more cells.

57. The method (1300) of any of claims 32 to 56, wherein an action is enabled onlywhen taking the action leaves enough time for the communication device (100) to reach the goal position without exceeding the maximum allowable time.

58. The method (1300) of any of claims 32 to 57, wherein the step of planning (S610,S1310) the trajectory further comprises:determining whether or not a candidate trajectory, which is output by theBayesian RL algorithm or the simulation-based MCTS algorithm, meets one or morerequirements.

59. The method (1300) of claim 58, further comprising:in response to determining that no candidate trajectory meets the one or morerequirements, reporting (S620) at least one of: -no trajectory can be planned provided the one or more requirements because ofno enough knowledge of the cellular environment; and -a level of knowledge of the cellular environment.

60. The method (1300) of any of claims 32 to 59, wherein after the step of planning(S610, S1310) the trajectory, the method (1300) further comprises at least one of:transmitting (S635), to the communication device (100), the planned trajectory;reporting (S630), to a traffic management system, the planned trajectory; andreceiving (S635), from the traffic management system, a message indicatingwhether the planned trajectory is allowed or rejected.

61. The method (1300) of claim 60, further comprising:re-planning a trajectory in response to receiving the message indicating that the planned trajectory is rejected.

62. The method (1300) of claim 60 or 61, further comprising:receiving (S635), from the communication device (100), a message indicating newRSRP measurements; and updating (S505) one or more per-cell datasets (^^) based on the new RSRP measurements.

63. A device (140, 1400) for planning a trajectory for a communication device (100),the device (140, 1400) comprising:a processor (1406);a memory (1408) storing instructions which, when executed by the processor(1406), cause the device (140, 1400) to performs the method of any of claims 32 to 62.

65. A computer program (1410) comprising instructions which, when executed by atleast one processor (1406), cause the at least one processor (1406) to carry out themethod (1200, 1300) of any of claims 1 to 29 and 32 to 62.

66. A carrier (1408) containing the computer program (1410) of claim 65, wherein thecarrier (1408) is one of an electronic signal, optical signal, radio signal, or computer readable storage medium.

67. A system (40) for trajectory planning, the system (40) comprising:one or more communication devices (100);a trajectory planning device (140) for planning a trajectory for at least one of thecommunication devices (100),wherein the trajectory planning device (140) comprises:a processor; a memory storing instructions which, when executed by the processor,cause the trajectory planning device (140) to:plan a trajectory for the at least one communication device (100) in a cellular environment based on at least a cellular world model that indicates a belief about what the cellular environment is.

68. The system (40) of claim 67, further comprising:a model generating device (145) for generating the cellular world model, thedevice (145) comprising:a processor; a memory storing instructions which, when executed by the processor,cause the model generating device (145) to:generate the cellular world model.

69. The system (40) of claim 68, wherein the instructions stored on memory of themodel generating device (145), when executed by the processor of the model generatingdevice (145), further cause the model generating device (145) to perform the method(1200) of any of claims 2 to 29.

70. The system (40) of any of claims 67 to 69, wherein the instructions stored on thememory of the trajectory planning device (140), when executed by the processor of thetrajectory planning device (140), further cause the trajectory planning device (140) toperform the method (1300) of any of claims 33 to 62.

Images