Systems and methods for drone delivery

Abstract

Systems, devices and methods generally related to unmanned aerial delivery vehicles are described herein. More specifically, systems, devices and methods related to improved, standardized, payloads and packaging are described. One example disclosed standardized packaging system comprises: a drone comprising a rectangular payload bay, wherein the payload bay has three dimensions, each based on a base unit of measure (b): a shortest side having a length of b; a medium side having a length of 2b; and a longest side having a length of 3b to 4b. In addition, a vertical take-off and landing vehicle and systems, devices and methods for identifying a balcony of a multi-story building in connection with completing a package delivery by an unmanned aerial vehicle are described.

Description

TITLE OF THE INVENTIONSYSTEMS AND METHODS FOR DRONE DELIVERYCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This patent application claims priority to U.S. Provisional Patent Application No. 63 / 655,100, entitled “Compact Low Noise Tailsitter Drone for Residential Package Delivery”, filed on June 3, 2024; and further claims priority to U.S. Provisional Patent Application No. 63 / 784,809, entitled “Drone Delivery and Landing System”, filed on April 7, 2025, the disclosure of each of which is hereby incorporated by reference as if set forth in their entirety herein.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] Not Applicable.NAMES OF THE PARTIES TO A JOINT RESEARCH AGREEMENT

[0003] Not Applicable.INCORPORATION BY REFERENCE OF MATERIAL SUBMITTED ON A COMPACT DISK

[0004] Not Applicable.BACKGROUND OF THE INVENTION1. Field

[0005] The subject matter disclosed herein generally relates to systems and methods for drone delivery. More particularly, the subject matter disclosed herein relates to unmanned-aircraft delivery systems and, in particular, to:• a modular packaging architecture that allows restaurant meals of diverse shapes and masses to be packed densely in a drone payload bay;• a computer-vision localisation technique for drone delivery to balconies.• a reconfigurable control- surface arrangement for convertible- wing (tailsitter) aircraft that reduces the vehicle’s vertical dimension while hovering, permitting flight inside the limited clearance typical of residential balconies.2. Background

[0006] Commercial food-delivery drones in service today operate almost exclusively in open, ground- level drop zones — for example backyards, parking lots, or suburban driveways. Three independent technical hurdles keep operators from addressing balconies and other elevated customer areas:1. Irregular restaurant packaging. Restaurants typically source clamshells, bowls, trays, and drink carriers from multiple suppliers. The resulting mix of footprints and heights wastes payload volume and forces staff to trial-and-error load plans, slowing kitchen throughput and encouraging use of oversized drones.2. Lack of balcony localisation methods. Global-navigation-satellite-system (GNSS) error in urban canyons can exceed the width of an apartment facade, and no widely adopted technique exists for guiding a drone to a particular balcony without mounting dedicated markers or relying on the resident to stand outside and signal.3. Vehicle geometry constraints. When a fixed-wing tailsitter transitions to hover, its wings and control surfaces stand vertically, producing a silhouette that can exceed the doorway-to-ceiling distance of many balconies. Existing fold-arm solutions suited to multirotor craft are not directly applicable to tailsitters, which must preserve clean aerodynamics in cruise.

[0007] Because these problems have not been jointly solved, present-day operators confine drops to open areas; elevated residential locations remain unserved.

[0008] To extend drone delivery to balconies, an integrated solution needs to do at least one of the following:• Standardise payload geometry so any combination of menu items packs efficiently, is centre -restrained automatically, and can be loaded in seconds by unskilled staff;• Locate individual balconies autonomously using information a resident can provide once, with an ordinary smartphone, and without installing external markers; and• Shrink the hovering footprint of a tailsitter far enough to enter the balcony volume while maintaining normal cruise performance elsewhere on the route.

[0009] The systems and methods for drone delivery described below address one or more of these objectives and collectively enable reliable, routine balcony delivery in environments where only backyard delivery was previously feasible.BRIEF SUMMARY OF EMBODIMENTS OF THE INVENTION

[0010] Accordingly, the present invention is directed to systems and methods for drone delivery that substantially obviates one or more problems resulting from the limitations and deficiencies of the related art.

[0011] In accordance with one or more embodiments of the present invention, there is provided a standardized packaging system for drone delivery that includes a drone comprising a rectangular payload bay, wherein the payload bay has three dimensions, each based on a base unit of measure (b): a shortest side having a length of b; a medium side having a length of 2b; and a longest side having a length of 3b to 4b.

[0012] In a further embodiment of the present invention, the longest side of the payload bay is aligned with a wingtip to wingtip direction of the drone.

[0013] In yet a further embodiment, the base unit of measure (b) is between 5 cm and 10 cm for a drone configured to deliver to balconies.

[0014] In still a further embodiment, the base unit of measure (b) is greater than 10 cm for a drone configured to deliver to open areas.

[0015] In yet a further embodiment, the standardized packaging system further comprises a plurality of food containers, wherein each food container has dimensions that are discrete multiples or discrete fractions of b, thereby efficiently utilizing space within the payload.

[0016] In still a further embodiment, the standardized packaging system further comprises a plurality of exterior packaging boxes [6] sized to fit within the payload bay, wherein each exterior packaging box comprises dimensions selected from the group consisting of: b by 2b by 2b; b by 2b by 3b; and b by 2b by 4b.

[0017] In yet a further embodiment, wherein the payload bay includes a mechanism configured to restrain and center exterior packaging boxes having a length that is less than the longest side of the payload bay.

[0018] In still a further embodiment, the mechanism comprises a plurality of flexible rods that are configured to apply force against a side of the exterior packaging and against a side wall of the payload.

[0019] In yet a further embodiment, the mechanism comprises electromechanical actuators configured to close gaps and center the exterior packaging boxes within the payload bay.

[0020] In still a further embodiment, a method for efficient drone delivery of food items, comprising: (i) providing a drone having a rectangular payload bay

[0010] wherein the payload bay has three dimensions based on a base unit of measure (b), the payload bay comprising dimensions of: b by 2b by 3b-4b; (ii) providing a plurality of food containers, each food container comprising dimensions that are discrete multiples or discrete fractions of b;(iii) packaging food items in the plurality of food containers; (iv) placing the plurality of food containers within an exterior packaging box; and (v) placing the exterior packaging box in the payload bay.

[0021] In yet a further embodiment, the exterior packaging box comprises dimensions selected from the group consisting of: b by 2b by 2b; b by 2b by 3b; and b by 2b by 4b.

[0022] In still a further embodiment, the method further comprises: restraining and centering one of the exterior packaging boxes within the payload bay using a restraining mechanism when the exterior packaging box comprises a length that is less than the longest dimension of the payload bay.

[0023] In accordance with one or more other embodiments of the present invention, there is provided a system for standardized food delivery via drones that includes a plurality of drones, each drone having a rectangular payload bay comprising dimensions based on a base unit of measure (b), each drone comprising a payload bay with dimensions of b by 2b by 4b, wherein the base unit is a predetermined length; a plurality of food containers having dimensions that are multiples or fractions of the base unit; and a plurality of exterior packaging boxes sized tocontain the food containers, wherein the exterior packaging boxes have dimensions selected from the group consisting of: b by 2b by 2b; b by 2b by 3b; and b by 2b by 4b.

[0024] In a further embodiment of the present invention, each drone comprises a restraining mechanism configured to center and secure exterior packaging boxes that comprise a length that is less than the longest dimension of the payload.

[0025] In yet a further embodiment, the food containers include containers of at least two different dimension sets, wherein each dimension set uses multiples or fractions of b for each side.

[0026] In still a further embodiment, the dimension b is between 5 cm and 10 cm, and the drones are configured to deliver to balconies.

[0027] In yet a further embodiment, the dimension b is greater than 10 cm, and the drones are configured to deliver to open areas.

[0028] In still a further embodiment, the longest dimension of the payload bay is aligned with a wingtip to wingtip direction of the drone, the medium dimension is aligned with a thrust direction of a propeller of the drone, and the shortest dimension is aligned with a remaining 3D axis, thereby enhancing pitch stability during wingborne flight when a payload's center of mass is off-centric.

[0029] In accordance with yet one or more other embodiments of the present invention, there is provided a method for identifying a balcony for drone delivery that includes (i) determining that a user resides in a multi-story building with a balcony; (ii) prompting the user to capture at least one image of a view from the balcony using a mobile device; (iii) transmitting the at least one image to a drone delivery system; (iv) navigating a drone to a vicinity of the multi-story building; (v) comparing, by the drone, a current image captured by a drone camera with the at least one image captured from the balcony; (vi) determining, using a relative camera pose estimation algorithm, coordinates of the balcony based on the comparison; and (vii) navigating the drone to the determined coordinates of the balcony.

[0030] In a further embodiment of the present invention, the step of capturing the at least one image comprises capturing a plurality of images and combining the plurality of images to form a panoramic image.

[0031] In yet a further embodiment, the relative camera pose estimation algorithm comprises a model with machine learnable parameters that receives both the at least one image captured from the balcony and the current image captured by the drone camera as inputs and outputs coordinates of the balcony within a frame of reference of the drone.

[0032] In still a further embodiment, the method further comprises: transmitting additional information from the mobile device, the additional location information comprising at least one of: GPS coordinates, address information, compass heading data, and barometric pressure data.

[0033] In yet a further embodiment, the method further comprises: processing the at least one image to extract image features; and transmitting only the extracted image features to the drone instead of the complete image.

[0034] In still a further embodiment, the method further comprises: comparing the at least one image with an aerial map of an area surrounding the multi-story building; determining a precise position and orientation of the balcony in global coordinates; and providing the precise position and orientation to the drone.

[0035] In yet a further embodiment, the method further comprises: refining the balcony coordinates during the first delivery; using sensors on the drone; and storing the balcony coordinates in a database associated with the user for use in future deliveries.

[0036] In still a further embodiment, determining that the user resides in a multi-story building with a balcony comprises at least one of: analyzing the user's address; analyzing GPS coordinates associated with the user; and receiving a selection from the user indicating a dwelling type.

[0037] In accordance with still one or more other embodiments of the present invention, there is provided a method for identifying a balcony for drone delivery that includes: (i) determining that a user resides in a multi-story building with a balcony; (ii) navigating a drone to a vicinity of the multi-story building based on approximate location information; (iii) prompting the user, via a mobile application, to capture an image of the drone when the drone is hovering near the multistory building; (iv) receiving the image of the drone from the user's mobile device;(v) determining, based on the image of the drone, a relative position between the balcony and the drone; and (vi) navigating the drone to the balcony based on the determined relative position.

[0038] In a further embodiment of the present invention, the step of determining the relative position between the balcony and the drone comprises: (i) identifying pixel coordinates of the drone within the image; (ii) obtaining orientation data of the mobile device when the image was captured; and (iii) calculating the relative position based on the pixel coordinates and the orientation data.

[0039] In still a further embodiment, determining the relative position between the balcony and the drone is done using a model with machine learnable parameters.

[0040] In yet a further embodiment, the approximate location information comprises at least one of: GPS coordinates associated with the user's address; barometric pressure data from the user’s mobile device; and floor number information provided by the user.

[0041] In still a further embodiment, the method further comprises: storing precise coordinates of the balcony for future deliveries after the drone has successfully navigated to the balcony.

[0042] In accordance with yet one or more other embodiments of the present invention, there is provided a system for identifying a balcony for drone delivery, comprising: (i) a server configured to: receive at least one image captured from a balcony of a user; store the at least one image in association with user information; and transmit the at least one image or features extracted from the at least one image to a drone; and (ii) a drone comprising: a camera; a processor; and a memory storing instructions; and whereby the instructions, when executed by the processor, cause the drone to: capture at least one current image of the scene in front of a multi-story building using the camera; compare the at least one current image with the at least one image captured from the balcony; determine, using a relative camera pose estimation algorithm, coordinates of the balcony based on the comparison; and navigate to the determined coordinates of the balcony.

[0043] In a further embodiment of the present invention, the drone is further configured to: fuse the coordinates determined using the relative camera pose estimation algorithm with additional location data to refine the balcony location, the additional location data comprising at least one of: three-dimensional GPS coordinates, barometric pressure readings, and floor count information.

[0044] In yet a further embodiment, the system further comprises a mobile application configured to: guide the user through a process of capturing the at least one image from the balcony; process the at least one image to extract features before transmission to the server; and capture environmental data from sensors of a mobile device.

[0045] In accordance with still one or more other embodiments of the present invention, there is provided a system for identifying a balcony for drone delivery, comprising a server; a drone; a mobile application configured to: (i) prompt a user to capture an image of a drone when the drone is hovering in a vicinity of a multi-story building; (ii) obtain orientation data of a mobile device when the image is captured; and (iii) transmit the image and the orientation data to a server; whereby the server and drone are configured to determine, based on the image and the orientation data, a relative position between a balcony and the drone.

[0046] In a further embodiment of the present invention, determining the relative position comprises: identifying pixel coordinates of the drone within the image; and calculating the relative position based on the pixel coordinates and the orientation data.

[0047] In yet a further embodiment, the drone is further configured to: store precise coordinates of the balcony after successfully navigating to the balcony; and use the precise coordinates for future deliveries to the balcony.

[0048] In still a further embodiment, the drone is initially navigated to the vicinity of the multi-story building based on approximate location information comprising at least one of: GPS coordinates associated with the user's address; barometric pressure data from the user's mobile device; and floor number information provided by the user.

[0049] In accordance with yet one or more other embodiments of the present invention, there is provided a vertical takeoff and landing (VTOL) aircraft comprising: a body; a plurality of control surfaces attached to the body; and actuators configured to move the control surfaces; wherein the actuators are configured to: position the control surfaces in a first configuration for forward flight or standard hovering flight; and position the control surfaces in a second configuration for reduced-height hovering flight by inverting at least a portion of the control surfaces, wherein the second configuration reduces at least one dimension of the aircraft compared to the first configuration.

[0050] In a further embodiment of the present invention, the aircraft is a tailsitter aircraft, and wherein the second configuration reduces the nose-to-tail length of the aircraft in hover mode compared to a nose-to-tail length of the aircraft in the first configuration.

[0051] In yet a further embodiment, the plurality of control surfaces comprises inner control surfaces and outer control surfaces; and in the second configuration: the inner control surfaces are rotated approximately 90 degrees relative to their position in the first configuration; and the outer control surfaces are rotated approximately 180 degrees relative to their position in the first configuration.

[0052] In still a further embodiment, the actuators for the outer control surfaces have a rotation range greater than 190 degrees to provide a control margin in both the first configuration and the second configuration.

[0053] In yet a further embodiment, the second configuration enables the aircraft to position itself closer to a delivery surface during package delivery compared to the first configuration.

[0054] In accordance with still one or more other embodiments of the present invention, there is provided a method for reducing at least one dimension of a vertical takeoff and landing (VTOL) aircraft for operation in confined spaces, comprising: reconfiguring control surfaces of the VTOL aircraft from a standard flight configuration to a reduced-height configuration by inverting or retracting at least a portion of the control surfaces; wherein the reconfigured control surfaces reduce at least one dimension of the VTOL aircraft while maintaining flight control capability during hovering.

[0055] In a further embodiment of the present invention, the VTOL aircraft comprises a tailsitter aircraft; the reconfiguration of the control surfaces reduces the vertical height of the tailsitter aircraft in hovering flight; and the reduced vertical height enables the tailsitter aircraft to enter spaces with limited vertical clearance.

[0056] It is to be understood that the foregoing general description and the following detailed description of the present invention are merely exemplary and explanatory in nature. As such, the foregoing general description and the following detailed description of the invention should not be construed to limit the scope of the appended claims in any sense.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0057] The invention will now be described, by way of example, with reference to the accompanying drawings, in which:

[0058] FIG. 1 depicts three different exterior boxes for the fractional standardized packaging system described herein, according to an illustrative embodiment;

[0059] FIG. 2 depicts food items packaged using a conventional packaging system (on left side of figure), compared to the fractional packaging system described herein (on right side of 1'igure), demonstrating increased space efficiency;

[0060] FIG. 3 depicts a drone with payload bay that includes flexible rods to center payloads with reduced size in the longest axis, according to an illustrative embodiment;

[0061] FIG. 4 depicts a drone payload bay with mechanical actuators to center payloads with reduced size in the longest axis, according to another illustrative embodiment;

[0062] FIG. 5 depicts a multi-story building with a user scanning the balcony view using their mobile device, along with a drone flying in front of the balcony and a remote server;

[0063] FIG. 6 depicts a smartphone application prompting the user to scan their balcony view with the camera, according to another illustrative embodiment;

[0064] FIG. 7 depicts a Siamese neural network used in some embodiments herein to compute the balcony location based on smartphone and drone camera images;

[0065] FIG. 8 depicts a drone in a standard configuration on the left side of the figure, and in an inverted configuration on the right side of the 1'igure, according to an illustrative embodiment; and

[0066] FIG. 9 depicts a side view of a drone in a standard configuration on the left side of the figure, and in an inverted configuration on the right side of the figure, according to an illustrative embodiment.

[0067] Throughout the figures, the same parts are always denoted using the same reference characters so that, as a general rule, they will only be described once.DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

[0068] As described hereinafter, embodiments of a drone delivery system that include a payload system, balcony identification, and tailsitter height reduction for balcony entry will be described.1. Payload System

[0069] In one or more embodiments described hereinafter, in order to increase the amount of useful payload that can be transported in a drone, particularly in a restaurant food delivery context, a standardized packaging system is described.

[0070] Similar to the way shipping containers have streamlined global trade, one or more embodiments of the present disclosure set forth a standardized packaging system for drone food delivery, allowing for greater delivery speed and lower delivery cost. In one embodiment, a drone comprises a rectangular payload bay, where the longest side is aligned with the wingtip to wingtip direction, and the second longest side is aligned with the thrust direction of the propeller, and the shortest side making up the remaining axis. If the payload’s center of mass is off-centric, this arrangement allows for much greater pitch stability during wingbome flight compared to the alternative option where the longest side is aligned with the flight direction. One challenge in making drone delivery boxes for restaurants is that as much density as possible is desired.Larger payload volume makes the drone larger, and therefore harder to navigate close to merchants and customers, and larger volume with the same weight leads to a slower optimal flight speed. Restaurants however, have a habit of using containers from restaurant packaging suppliers with little concern for volumetric efficiency. What is worse is that a container may be just a little bit too long on one side for the drone payload box. For example, one item might have a footprint of 19 x 7cm, and another item 12 x 12cm. These two items cannot fit inside of a payload box with a footprint of 16 x 32cm, because the first item only fits sideways, blocking the space for the second item. To solve this problem, in one or more embodiments, a fractional standardized packaging system was developed comprising two important elements: (i) a rectangular drone payload bay with dimensions based upon multiples of base units; and (ii) restaurants and their packaging suppliers being instructed or incentivized to create packaging material where every side is either a multiple or a fraction of the base unit.

[0071] With regard to the first element of the packaging system, in these one or more embodiments, the payload bay is rectangular and has a shortest side with a length of 1 base unit, a long side with a length of 3 or 4 base units, and a medium side with a length of 2 base units. The base unit should be between 5 cm and 10 cm for a drone that can deliver to balconies, or more than 10 cm for drones that can only deliver to open areas such as backyards.

[0072] With regard to the second element of the packaging system, in these one or more embodiments, restaurants and their packaging suppliers create packaging material where every side is either a multiple or a fraction of the base unit. For example, a box for hamburgers may be dimensioned lb* lb* lb, where lb denotes one base unit. This hamburger box allows eight (8) hamburgers to be placed within a payload bay of 1x2x4 base units. And because the three (3) dimensions of the payload bay are discrete multiples of the base unit, a low skilled worker at the restaurant has infinite ways to fit those eight (8) boxes. The worker does not have to orient them in a specific way, due to the universality of the base unit. The worker can rotate every box by 90 degrees and it still fits. Another piece of packaging material might be a plastic container which is 2 base units long, 1 base unit wide, and half a base unit tall. Such a box may be useful for sushi, naan bread, or a sandwich. Suppose an order contains 2 portions of sushi, and 5 burgers. This can also fit into the standardized drone delivery box easily, and here again, the worker has a large number of different ways to position and orient the boxes. The sushi boxes can be placed vertically or horizontally, but in either direction, they use their space optimally, leaving enough space for the 5 burger boxes. This portfolio of packaging materials can be extended to beverage containers, sauce containers, etc. A large number of different containers are possible, and as long as they are rectangular, with each side having a length that is a multiple or a fraction of the base unit, it will fit easily into a payload box. In addition, the payload bay of the drone can be built in such a way that it can accept payloaded boxes of three (3) different dimensions: lb*2b*2b, lb*2b*3b and lb*2b*4b. The two (2) shorter sides are fixed, but the longest side can be variable. For example, referring to FIG. 1, exterior packaging boxes 104 having three (3) different dimensions (lb*2b*2b, lb*2b*3b and lb*2b*4b) are shown. In FIG. 1, each exterior packaging box 104 has a shortest side 101 with length lb, a medium side 102 with length 2b, and a longest side with length 2b, 3b, or 4b. To enable the acceptance of these different sized boxes, the payload bay has a dimension of lb*2b*4b, but also has a mechanism on the side that restrains smaller payloads and fixes them in the center of the payload bay. This mechanism can either be passive or active. A passive mechanism may comprise two (2) flexible rods 307 on each side of a rectangular drone payload bay 305 that press inward (refer to FIG. 3), an active mechanism may comprise an electromechanical actuator 401 on each side of a payload bay of a drone 404 that actively closes the gap and / or pushes the payload box towards the center (see FIG. 4), combined with a set of sensors or cameras, and a computersystem which commands the actuators to match the width of the particular payload. In this embodiment, the restaurants are mandated that all their exterior packaging boxes must be either lb*2b*2b or lb*2b*3b or lb*2b*4b. This gives flexibility, because the restaurant can choose from three (3) different exterior boxes, but they all fit into the payload bay, because the two short sides match the payload bay dimension, and the third side is flexible due to the mechanism that holds the box into the middle. Also note that each of these three (3) exterior boxes is compatible with the fractional packaging system. This means that all the sub-containers fit easily, in any orientation. For example, FIG. 2 depicts food items packaged using a conventional packaging system (on left side of FIG. 2), compared to the fractional packaging system described herein (on right side of FIG. 2), demonstrating increased space efficiency. On the right side of FIG. 2, a plurality of food containers 201 are arranged in three (3) different possible configuration according to the fractional packaging system described herein.2. Balcony Identification

[0073] When delivering a package using a drone, finding the right balcony can be a challenge. In one or more embodiments described hereinafter, a unique method for finding the right balcony is disclosed, hi accordance with this method, when a user first sets up his or her account for drone delivery, it is determined whether or not he lives in a single family home, high rise apartment building, or other.

[0074] In one or more embodiments, the dwelling type determination can be made either through his / her address, Global Positioning System (GPS) information, or the user can be asked to select which dwelling type describes him best. If it has been determined that the user lives in an apartment with a balcony, and no access to a yard (or access to a yard but preference for balcony delivery), then the user is prompted to go on his / her balcony with their smartphone, and scan the view using their smartphone camera. For example, FIG. 5 depicts a multi-story building 501 with a user 506 scanning the view from his or her balcony 502 using his or her mobile device 503. FIG. 5 additionally illustrates a drone 404 with a drone camera 504 flying in front of the balcony 502. A remote server 505 operatively coupled to the mobile device 503 of the user 506 and the drone 404 is also diagrammatically illustrated in FIG. 5. FIG. 6 depicts an an example of a smartphone application 601 prompting the user to take a picture of the view from his or her balcony. The user-captured image from the balcony can take the form of taking asimple picture, or it can take the form of a short video, or it can be a guided scanning process similar to the “panorama picture” function found in most default smartphone camera applications (“apps”). It is advisable to use one of the later methods, because then a larger field of view can be captured. When the video or series of pictures are taken, they can be stitched together using common image stitching algorithms. The result of this stitching process is a wide angle image. However, as stated earlier, a single image may be sufficient, particularly if the phone has a wide angle camera, or an unstitched collective of images may be used for the next steps, but a stitched wide angle image may be the easiest solution. Once that balcony image is obtained, in one embodiment, it is then sent to the server of the drone delivery operator, from where it is then sent to the drone. In these one or more embodiments, other location information such as GPS information (e.g., GPS coordinates), address, compass heading and smartphone barometer pressure may be sent as well.

[0075] In these one or more embodiments, once the drone arrives at the customer’s house, the drone runs a relative-camera pose estimation algorithm, comparing the user’s balcony image with the view from the drone camera, and generating the relative position of the drone and user camera / balcony. This relative camera pose estimator is best implemented using a deep learning algorithm, where a neural network receives both images as input (the users balcony view, and the drones camera) and the output arc the coordinates of the users camera, within the drones reference frame, giving the drone knowledge of the user’s balcony location. For the neural network, a so-called Siamese convolutional neural network (CNN) may be used, but other architectures could work too. Someone skilled in the art can refer to the papers that have been published on relative camera pose estimation to create such a model or algorithm. In the case of a deep learning based solution, it is advised to generate a dataset of at least ten-thousand balcony-drone image pairs with coordinates, to train such models on. To accomplish this, synthetic simulator datasets may be easiest. There are commercial service providers who offer custom synthetic datasets. It may be useful or necessary to generate a dataset that mimics the cameras of the drone. Generally speaking, wide angle cameras are recommended for the drone, because this way there is a larger overlap between the features captured by the drone and by the user’s smartphone. As anyone skilled in the art knows, it may be better to use a neural network pretrained on another computer vision task such as image classification, and then fine-tuned on the balcony identification task, such that a smaller dataset can be used, compared to a modeltrained from scratch. If issues are encountered with training the model from a synthetic dataset, commonly known techniques for “sim to real” transfer may be used. A real life dataset may be recorded as well, or instead of a synthetic dataset. To do this, one of the easiest ways is to fly a drone in front of a house facade, capturing images from the perspective of the balcony, and then generating a set of quasi balcony pictures. Then later, a drone can be flown in front of the building, in various locations. The images recorded from the drones perspective can then be labeled manually. To do this, a simple user interface can be created, where a quasi-balcony picture is displayed together with a drone picture, and then a dataset engineer can click on the drone picture to identify the location from which the balcony picture was taken. Other methods may be used, for example, one can record a dataset in a miniature park, where one can have a small camera, or small set of cameras attached to a stick, and hold the stick in front of miniature buildings, simulating balcony images and drone images. Once a dataset of sufficient size and quality is obtained, anyone skilled in the ail can train the model for the task.

[0076] In these one or more embodiments, the output of this model does not have to be just the camera location that the balcony image was taken from. It could also include more sophisticated information, such as the coordinates of the center of the balcony, or coordinates defining a bounding box of the balcony. A neural network can learn such information because it is a solvable problem given the drone’s camera images. A neural network or equivalent can learn to see where the balcony image was taken from, and then size up and locate the actual balcony.

[0077] In these one or more embodiments, the drone, as it hovers in front of the building, runs both the on board camera images, and the users balcony image through the computer vision module, and obtains the coordinates from which the camera image was taken. This tells the drone the target balcony location, and the drone’ s autopilot system can then fly to the balcony and perform the delivery.

[0078] In these one or more embodiments, it should be noted that the user’s smartphone app may not need to send the actual panorama image file to the server, and the server may not need to send the image file to the drone. Instead, a compressed set of image features may be sent over. For example, in the case of a Siamese deep neural network (Siamese meaning two (2) image input branches), the balcony image branch of the neural network may be ran on the phone or on the server, and then only the output vector of the balcony branch can be sent to the drone. And then the drone only processes the remaining parts of the neural network. In addition to therelative camera position estimation process on the drone, the balcony image may also be used on the server, to compare it with an aerial map of the area, to determine the precise position and orientation of the user’s balcony in global coordinates. Similar to the Siamese neural network with a balcony branch and a drone branch, where the output is the relative position between the two, a similar architecture can be used, where the model has a balcony branch and an aerial map branch, normed to a certain scale, and centered around a certain coordinate point. And then, the output of the Siamese neural network is the position and orientation of the user’s camera on that aerial map. This information can then be used in a sensor fusion algorithm such as a Kalman filter, to combine it with the smartphone’s GPS location and / or the GPS location associated with the user’s address. Even in the absence of an aerial image, such a server side model that processes the balcony image may estimate the altitude above ground from which the picture was taken. This can similarly augment the information that has been obtained on the user’s GPS point location and address. FIG. 7 depicts an example of a Siamese neural network used in some embodiments herein to compute the balcony location based on smartphone and drone camera images.

[0079] In tall high rises, where the balcony is many floors above the ground, the dronebalcony relative camera pose estimation may have a few meters of uncertainty, making it insufficient to reliably identify the correct balcony. In such cases, the balcony location predicted by the relative camera pose estimation may be fused with other estimates, such as the three dimensional GPS location of the user’s smartphone, or the barometer of the user’s smartphone. Barometer pressure can only be converted to altitude, if the local weather related pressure is known. In case the user smartphone barometer reading is not taken at the same time the drone is hovering outside the building, weather data may be used to adjust the barometer altitude. Such information may be fused with a Kalman filter or similar. Smartphone operating systems, such as IOS and android, may provide smartphone location data through their software development kit (SDK) that is already the result of sensor fusion. Such fused location data may be read from the SDK or system application programming interface (API) and used to refine the balcony position estimate. The computer vision system on the drone may also be configured to count the number of floors on a building visually. This floor count can then be compared to the floor number that the user specifies during the sign up process.

[0080] In these one or more embodiments, if the method(s) described above are still not reliable enough, another method may be used to confirm the balcony. According to this method, when the drone is hovering outside (note that this does not have to be the delivery drone, it can be a drone dedicated for the balcony identification process) then the user is prompted to step on the balcony again, and take a picture of the drone (it may not be presented as “taking a picture”, on the user interface, it may simply appear as scanning the view with their camera). Once a picture of the drone, from the user’s perspective, is obtained this can be used to determine the relative offset between the balcony and the drone. This may be done again with a deep neural network, trained on images of a view with a drone in it, where the output is the drone’s position, from within the images reference frame. Alternatively, it may be done with a simpler computer vision technique that merely tells us the drone’s pixel position, which may then be combined with the smartphone’ s orientation (which can be read from the smartphone system, which provides attitude / orientation using the gyroscope, accelerometer and / or other sensors). When the pixel location of the drone is known, as well as the orientation of the smartphone when the picture was taken, one can calculate the relative position of the drone and the camera in 3D space. This information can then be sent to the drone (directly or indirectly) telling the drone which balcony to deliver to. Again, this drone may do a delivery, or may not do a delivery and simply do the balcony setup process. Because once one drone has correctly identified the balcony, it can store the balcony location, in global coordinates, with enough precision that future deliveries do not need to do such computer vision processes anymore to do the delivery. If the initial setup drone (or first delivery drone) is equipped with real-time kinetic (RTK) enhanced GPS, or other accurate sensors that help positioning, such as visual odometry, optical flow, etc. Then the drone position in global coordinates can be known with high accuracy, and so during the first delivery or during the setup flight, the balcony location can be stored with accurate global coordinates as well.3. Tailsitter Height Reduction for Balcony Entry

[0081] In one or more embodiments described hereinafter, tailsitter height reduction systems and methods for balcony entry are disclosed. With regard to these balcony finder and balcony entry systems and methods, it is noted that these aforedescribed systems and methods can be used with various drone form factors, and are not restricted to tailsitters. However, when atailsitter is used, there is a technique that allows the tailsitter to reduce its effective height, which makes it easier to enter balconies, and also allows the drone to get closer to the delivery surface when the delivery is performed.

[0082] As can be seen in FIG. 8, the drone is configured such that the flaps on the body and on the side of the body can be actuated separately, and then the body flaps are rotated roughly 90 degrees (or more), and the outer flaps are rotated roughly 180 degrees, inverting them in a way where they can still control the yaw and pitch of the drone, but make the drone shorter vertically. For the outer flaps, actuators should be chosen with a range greater than 200 degrees, such that enough margin exists both in normal mode and inverted mode to move the flap back and forth to control the drone. For the body flaps, less range is needed, because the body flaps cannot invert completely, because that would make them intersect with the drone body. Instead, the body flaps may be controlled with a standard 180 degree range going from -90 to +90, and then during the height reduction procedure where the outer flaps are inverted, the body flaps can be set (close) to one of these maximum values of +90 or -90, making them horizontal. Obviously the drone cannot fly in cruise mode (wingbome flight) in the inverted configuration. The inverted configuration can only be used in hover mode (rotorcraft flight). Flap inversion may be initiated right after transition from cruise to hover mode, or before balcony entrance. The flap inversion may then be reversed, after exiting the balcony, or before transitioning back from hover to cruise mode. For example, as shown in FIG. 8, a vertical takeoff and landing (VTOL) aircraft or drone 801 is depicted in a first standard configuration 805 on the left side of FIG. 8, and in a second inverted configuration 806 on the right side of FIG. 8 for reduced height. In FIG. 8, the VTOL aircraft 801 has a body 802 and a plurality of control surfaces or flaps 803 that arc driven by a plurality of actuators 804. FIG. 9 depicts a side view of the VTOL aircraft or drone in a standard configuration 805 on the left side of FIG. 9, and in an inverted configuration on the right side of FIG. 9.

[0083] Any of the features or attributes of the above described embodiments and variations can be used in combination with any of the other features and attributes of the above described embodiments and variations as desired.

[0084] Although the invention has been shown and described with respect to a certain embodiment or embodiments, it is apparent that this invention can be embodied in manydifferent forms and that many other modifications and variations are possible without departing from the spirit and scope of this invention.

[0085] Moreover, while exemplary embodiments have been described herein, one of ordinary skill in the art will readily appreciate that the exemplary embodiments set forth above are merely illustrative in nature and should not be construed as to limit the claims in any manner. Rather, the scope of the invention is defined only by the appended claims and their equivalents, and not, by the preceding description.

[0086] The invention claimed is:

Claims

1. A standardized packaging system for drone delivery, comprising: a drone comprising a rectangular payload bay, wherein the payload bay has three dimensions, each based on a base unit of measure (b): a shortest side having a length of b; a medium side having a length of 2b; and a longest side having a length of 3b to 4b.

2. The standardized packaging system of claim 1, wherein the longest side is aligned with a wingtip to wingtip direction of the drone.

3. The standardized packaging system of claim 1, wherein b is between 5 cm and 10 cm for a drone configured to deliver to balconies.

4. The standardized packaging system of claim 1, wherein b is greater than 10 cm for a drone configured to deliver to open areas.

5. The standardized packaging system of claim 1, further comprising: a plurality of food containers, wherein each food container has dimensions that arc discrete multiples or discrete fractions of b, thereby efficiently utilizing space within the payload.

6. The standardized packaging system of claim 1, further comprising: a plurality of exterior packaging boxes sized to fit within the payload bay, wherein each exterior packaging box comprises dimensions selected from the group consisting of: b by 2b by 2b; b by 2b by 3b; and b by 2b by 4b.

7. The standardized packaging system of claim 6, wherein the payload bay includes a mechanism configured to restrain and center exterior packaging boxes having a length that is less than the longest side of the payload bay.

8. The standardized packaging system of claim 7, wherein the mechanism comprises a plurality of flexible rods that are configured to apply force against a side of the exterior packaging and against a side wall of the payload.

9. The standardized packaging system of claim 7, wherein the mechanism comprises electromechanical actuators configured to close gaps and center the exterior packaging boxes within the payload bay.

10. A method for efficient drone delivery of food items, comprising: providing a drone having a rectangular payload bay wherein the payload bay has three dimensions based on a base unit of measure (b), the payload bay comprising dimensions of: b by 2b by 3b-4b; providing a plurality of food containers, each food container comprising dimensions that are discrete multiples or discrete fractions of b; packaging food items in the plurality of food containers; placing the plurality of food containers within an exterior packaging box; and placing the exterior packaging box in the payload bay.

11. The method of claim 10, wherein the exterior packaging box comprises dimensions selected from the group consisting of: b by 2b by 2b; b by 2b by 3b; and b by 2b by 4b.

12. The method of claim 10, further comprising: restraining and centering one of the exterior packaging boxes within the payload bay using a restraining mechanism when the exterior packaging box comprises a length that is less than the longest dimension of the payload bay.- l-13. A system for standardized food delivery via drones, comprising: a plurality of drones, each drone having a rectangular payload bay comprising dimensions based on a base unit of measure (b), each drone comprising a payload bay with dimensions of b by 2b by 4b, wherein the base unit is a predetermined length; a plurality of food containers having dimensions that are multiples or fractions of the base unit; and a plurality of exterior packaging boxes sized to contain the food containers, wherein the exterior packaging boxes have dimensions selected from the group consisting of: b by 2b by 2b; b by 2b by 3b; and b by 2b by 4b.

14. The system of claim 13, wherein each drone comprises a restraining mechanism configured to center and secure exterior packaging boxes that comprise a length that is less than the longest dimension of the payload.

15. The system of claim 13, wherein the food containers include containers of at least two different dimension sets, wherein each dimension set uses multiples or fractions of b for each side.

16. The system of claim 13, wherein b is between 5 cm and 10 cm, and the drones are configured to deliver to balconies.

17. The system of claim 13, wherein b is greater than 10 cm, and the drones are configured to deliver to open areas.

18. The system of claim 13, wherein the longest dimension of the payload bay is aligned with a wingtip to wingtip direction of the drone, the medium dimension is aligned with a thrust direction of a propeller of the drone, and the shortest dimension is aligned with a remaining 3D axis, thereby enhancing pitch stability during wingbome flight when a payload's center of mass is off-centric.

19. A method for identifying a balcony for drone delivery, comprising: determining that a user resides in a multi-story building with a balcony; prompting the user to capture at least one image of a view from the balcony using a mobile device; transmitting the at least one image to a drone delivery system; navigating a drone to a vicinity of the multi-story building; comparing, by the drone, a current image captured by a drone camera with the at least one image captured from the balcony; determining, using a relative camera pose estimation algorithm, coordinates of the balcony based on the comparison; and navigating the drone to the determined coordinates of the balcony.

20. The method of claim 19, wherein capturing the at least one image comprises capturing a plurality of images and combining the plurality of images to form a panoramic image.

21. The method of claim 19, wherein the relative camera pose estimation algorithm comprises a model with machine learnable parameters that receives both the at least one image captured from the balcony and the current image captured by the drone camera as inputs and outputs coordinates of the balcony within a frame of reference of the drone.

22. The method of claim 19, further comprising: transmitting additional information from the mobile device, the additional location information comprising at least one of: GPS coordinates, address information, compass heading data, and barometric pressure data.

23. The method of claim 19, further comprising: processing the at least one image to extract image features; and transmitting only the extracted image features to the drone instead of the complete image.

24. The method of claim 19, further comprising: comparing the at least one image with an aerial map of an area surrounding the multi-story building; determining a precise position andorientation of the balcony in global coordinates; and providing the precise position and orientation to the drone.

25. The method of claim 19, further comprising: refining the balcony coordinates during the first delivery; using sensors on the drone; and storing the balcony coordinates in a database associated with the user for use in future deliveries.

26. The method of claim 19, wherein determining that the user resides in a multi-story building with a balcony comprises at least one of: analyzing the user's address; analyzing GPS coordinates associated with the user; and receiving a selection from the user indicating a dwelling type.

27. A method for identifying a balcony for drone delivery, comprising: determining that a user resides in a multi-story building with a balcony; navigating a drone to a vicinity of the multi-story building based on approximate location information; prompting the user, via a mobile application, to capture an image of the drone when the drone is hovering near the multi-story building; receiving the image of the drone from the user's mobile device; determining, based on the image of the drone, a relative position between the balcony and the drone; and navigating the drone to the balcony based on the determined relative position.

28. The method of claim 27, wherein determining the relative position between the balcony and the drone comprises: identifying pixel coordinates of the drone within the image; obtaining orientation data of the mobile device when the image was captured; and calculating the relative position based on the pixel coordinates and the orientation data.

29. The method of claim 27, wherein determining the relative position between the balcony and the drone is done using a model with machine learnable parameters.

30. The method of claim 27, wherein the approximate location information comprises at least one of: GPS coordinates associated with the user's address; barometric pressure data from the user's mobile device; and floor number information provided by the user.

31. The method of claim 28, further comprising: storing precise coordinates of the balcony for future deliveries after the drone has successfully navigated to the balcony.

32. A system for identifying a balcony for drone delivery, comprising: a server configured to: receive at least one image captured from a balcony of a user; store the at least one image in association with user information; and transmit the at least one image or features extracted from the at least one image to a drone; and a drone comprising: a camera; a processor; and a memory storing instructions; and whereby the instructions, when executed by the processor, cause the drone to: capture at least one current image of the scene in front of a multi- story building using the camera; compare the at least one current image with the at least one image captured from the balcony; determine, using a relative camera pose estimation algorithm, coordinates of the balcony based on the comparison; and navigate to the determined coordinates of the balcony.

33. The system of claim 32, wherein the drone is further configured to: fuse the coordinates determined using the relative camera pose estimation algorithm with additional location data to refine the balcony location, the additional location data comprising at least one of: three-dimensional GPS coordinates, barometric pressure readings, and floor count information.

34. The system of claim 32, further comprising a mobile application configured to: guide the user through a process of capturing the at least one image from the balcony; process the at least one image to extract features before transmission to the server; and capture environmental data from sensors of a mobile device.

35. A system for identifying a balcony for drone delivery, comprising: a server; a drone; a mobile application configured to: prompt a user to capture an image of a drone when the drone is hovering in a vicinity of a multi-story building; obtain orientation data of a mobile device when the image is captured; and transmit the image and the orientation data to a server; whereby the server and drone are configured to determine, based on the image and the orientation data, a relative position between a balcony and the drone.

36. The system of claim 35, wherein determining the relative position comprises: identifying pixel coordinates of the drone within the image; and calculating the relative position based on the pixel coordinates and the orientation data.

37. The system of claim 35, wherein the drone is further configured to: store precise coordinates of the balcony after successfully navigating to the balcony; and use the precise coordinates for future deliveries to the balcony.

38. The system of claim 35, wherein the drone is initially navigated to the vicinity of the multi-story building based on approximate location information comprising at least one of: GPS coordinates associated with the user's address; barometric pressure data from the user's mobile device; and floor number information provided by the user.

39. A vertical takeoff and landing (VTOL) aircraft comprising: a body; a plurality of control surfaces attached to the body; and actuators configured to move the control surfaces; wherein the actuators are configured to: position the control surfaces in a first configuration for forward flight or standard hovering flight; and position the control surfaces in a second configuration for reduced-height hovering flight by inverting at least a portion of the control surfaces, wherein the second configuration reduces at least one dimension of the aircraft compared to the first configuration.

40. The VTOL aircraft of claim 39, wherein the aircraft is a tailsitter aircraft, and wherein the second configuration reduces the nose-to-tail length of the aircraft in hover mode compared to a nose-to-tail length of the aircraft in the first configuration.

41. The VTOL aircraft of claim 39, wherein: the plurality of control surfaces comprises inner control surfaces and outer control surfaces; and in the second configuration: the inner control surfaces are rotated approximately 90 degrees relative to their position in the first configuration; and the outer control surfaces are rotated approximately 180 degrees relative to their position in the first configuration.

42. The VTOL aircraft of claim 41, wherein the actuators for the outer control surfaces have a rotation range greater than 190 degrees to provide a control margin in both the first configuration and the second configuration.

43. The VTOL aircraft of claim 39, wherein the second configuration enables the aircraft to position itself closer to a delivery surface during package delivery compared to the first configuration.

44. A method for reducing at least one dimension of a vertical takeoff and landing (VTOL) aircraft for operation in confined spaces, comprising: reconfiguring control surfaces of the VTOL aircraft from a standard flight configuration to a reduced-height configuration by inverting or retracting at least a portion of the control surfaces; wherein the reconfigured control surfaces reduce at least one dimension of the VTOL aircraft while maintaining flight control capability during hovering.

45. The method of claim 44, wherein: the VTOL aircraft comprises a tailsitter aircraft; the reconfiguration of the control surfaces reduces the vertical height of the tailsitter aircraft in hovering flight; and the reduced vertical height enables the tailsitter aircraft to enter spaces with limited vertical clearance.-SO-

Images