Object transfer system and method

Abstract

An object transfer system configured to transfer an object between autonomous aerial vehicles. The system includes an object transfer mechanism configured to transfer an object between the vehicles, a pose detection system configured to detect a relative pose of the vehicles and one or more processing devices configured to use the pose detection system to determine a relative pose of the vehicles, control at least one of the vehicles in accordance with the determined relative pose so that the vehicles have a target relative pose and operate the object transfer mechanism to thereby transfer the object between the vehicles.

Description

OBJECT TRANSFER SYSTEM AND METHODBackground of the Invention

[0001] The present invention relates to an object transfer system and method for transferring an object between autonomous aerial vehicles (UAVs).Description of the Prior Art

[0002] The reference in this specification to any prior publication (or information derived from it), or to any matter which is known, is not, and should not be taken as an acknowledgement or admission or any form of suggestion that the prior publication (or information derived from it) or known matter forms part of the common general knowledge in the field of endeavour to which this specification relates.

[0003] Swarm drones have opened up new frontiers in aerial technology, allowing for unique capabilities and applications. Inspired by the natural occurrence of collective behavior in animals, these autonomous cooperative flying robots offer immense potential in various fields. Multi-rotor Vertical Take-Off and Landing (VTOL) unmanned aerial vehicles (UAVs) are widely utilized drones, primarily due to their hovering capabilities and suitability for operating in confined spaces. From gas sensing to navigate cluttered environments autonomously, swarm drones demonstrate the power of collective intelligence and efficient cooperation. Addressing the power limitations is crucial to enable continuous sensing capability to a swarm of drones.

[0004] Research on modern LiPo batteries used in drones has come a long way in size and efficiency. Even though the efforts to improve efficiency are significant, the constraints of finite flight time persist. There are various methods developed to increase their efficiency. Various categories of drones have been developed to enhance their performance, and they can be classified into two distinct approaches. The first approach focuses on onboard techniques to improve flight efficiency. The second approach involves replenishing the energy source of the drone using different techniques such as autonomous or human-assisted ground-based battery swapping, see for example Guetta, Y. and Shapiro, A. "On-Board Physical Battery Replacement System and Procedure for Drones During Flight", IEEE Robotics and Automation Letters. 7, 4 (Oct. 2022), 9755-9762.

[0005] However, these methods have several disadvantages and limit the continuous operation of drones, cause significant logistics and time delays, and require complex ground stations equipped with swapping mechanisms which might be cumbersome in most cases.

[0006] One approach to this challenge involves utilizing a secondary drone that docks onto a primary drone, providing an additional energy source to extend its flight duration, as described for example in Jain, K.P. and Mueller, M.W. "Flying batteries: In-flight battery switching to increase multirotor flight time", Proceedings - IEEE International Conference on Robotics and Automation. (Sep. 2019), 3510-3516. However, a practical implementation encounters challenges such as ensuring reliable docking and undocking maneuvers and downwash effects, maintaining the extra weight of the secondary drone, and considering the impact of increased complexity on system reliability. Such issues are discussed in Jain, K.P. et al. "Docking two multirotors in midair using relative vision measurements", (Nov. 2020).

[0007] A. Shankar, S. Elbaum and C. Detweiler, "Towards In-Flight Transfer of Payloads Between Multirotors," in IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 6201-6208, Oct. 2020, describes an approach for two multirotors to transfer a payload between them inair, while keeping the payload aloft and stationary. The framework is built on a visual-feedback and grasping pipeline that enables one UAS to grasp the payload held by another, thereby allowing the UASs to act as swappable carriers. By connecting the payload outwards along a single rigid link, and allowing the UASs to maneuver about it, the pay load remains online while it is transferred to a different carrier. Furthermore, building entirely on monocular vision, the approach does not rely on precise extrinsic localization systems.

[0008] US9043052 describes a system and method for controlling a plurality of vehicles to affect positioning of a common payload. The system comprises of multiple vehicles having positioners to change the location of the common payload, where the group of vehicles form a swarm that is controlled by a driver or pilot station. Each vehicle is autonomously stabilized and guided through a swarm electronics unit, which further includes sensor, communication, and processing hardware. At the driver or pilot station, a system or a person remotely enters payload destinations, which is processed and communicated to each vehicle. The method for controlling a multi-vehicle system includes inputting the desired location of the payload and determining a series of intermediary payload waypoints. Next, these payload waypoints areused by the swarm waypoint controller to generate individual waypoints for each vehicle. A controller for each vehicle moves the vehicle to these individual waypoints.

[0009] Mellinger, D., Shomin, M., Michael, N., Kumar, V. (2013), "Cooperative Grasping and Transport Using Multiple Quadrotors". Distributed Autonomous Robotic Systems, Springer Tracts in Advanced Robotics, vol 83. Springer, Berlin, Heidelberg, describes controlling multiple quadrotor robots that cooperatively grasp and transport a pay load in three dimensions. The paper models the quadrotors both individually and as a group rigidly attached to a payload. The paper proposes individual robot control laws defined with respect to the payload that stabilize the payload along three-dimensional trajectories. The paper details the design of a gripping mechanism attached to each quadrotor that permits autonomous grasping of the payload.Summary of the Present Invention

[0010] In one broad form, an aspect of the present invention seeks to provide an object transfer system configured to transfer an object between autonomous aerial vehicles, the system including: an object transfer mechanism configured to transfer an object between the vehicles; a pose detection system configured to detect a relative pose of the vehicles; and, one or more processing devices configured to: use the pose detection system to determine a relative pose of the vehicles; control at least one of the vehicles in accordance with the determined relative pose so that the vehicles have a target relative pose; and, operate the object transfer mechanism to thereby transfer the object between the vehicles.

[0011] In one embodiment the pose detection system includes: an imaging device mounted on one of the vehicles; and, a visual marker mounted on an other one of the vehicles, and wherein the one or more processing devices are configured to determine a relative pose of the vehicles by: using the imaging device to capture images of the visual marker; and, analysing the images to determine a relative pose of the first and second vehicles.

[0012] In one embodiment the one or more processing devices are configured to determine a relative pose of the vehicles based on a geometry of the visual marker in the captured images.

[0013] In one embodiment the visual marker includes machine readable coded data and wherein the one or more processing devices are configured to determine the relative pose at least in part using the coded data.

[0014] In one embodiment the coded data is indicative of a visual marker identity and wherein the one or more processing devices are configured to determine the relative pose at least in part using the visual marker identity.

[0015] In one embodiment the visual marker includes multiple visual markers and wherein the one or more processing devices are configured to determine the relative pose at least in part based on at least one of: relative sizes of the multiple visual markers; relative positions of the multiple visual markers; relative orientations of the multiple visual markers; and, identities of the multiple visual markers.

[0016] In one embodiment the object transfer mechanism includes: a deployment apparatus mounted on a first one of the vehicles; and, a receiver apparatus mounted on a second one of the vehicles.

[0017] In one embodiment the object deployment apparatus includes a deployment member configured to transfer the object to the receiver apparatus.

[0018] In one embodiment the deployment member is configured selectively engage the object receiver apparatus.

[0019] In one embodiment the deployment member is configured to magnetically engage the object receiver apparatus.

[0020] In one embodiment the deployment member includes a slide configured to slidably deploy the object.

[0021] In one embodiment the deployment apparatus includes a deployment actuator configured to move the deployment member between extended and retracted positions.

[0022] In one embodiment the deployment apparatus includes: a housing configured to retain the object; and, a release actuator configured to release the object from the housing.

[0023] In one embodiment the housing includes multiple slots, each being configured to retain an object and wherein the one or more processing devices are configured to control the release actuator to thereby release the object from one of the slots.

[0024] In one embodiment the object receiver apparatus includes a receiver member configured to receive the object from the deployment apparatus.

[0025] In one embodiment the receiver member includes a chute.

[0026] In one embodiment the receiver apparatus includes a receiver actuator configured to move the receiver member between extended and retracted positions.

[0027] In one embodiment the one or more processing devices are configured to: determine once the first and second vehicles have the target relative pose; operate deployment and receiver actuators so that deployment and receiver members extend and engage; and, operate a release actuator to release the object from a housing once the deployment and receiver members are engaged to thereby transfer the object from the first vehicle to the second vehicle.

[0028] In one embodiment the target relative pose is configured at least one of: so that a first vehicle is positioned at least partially above and laterally offset from a second vehicle; and, to minimise an impact of downwash from the first vehicle.

[0029] In one embodiment the object transfer mechanism includes at least one of: a robot arm including an end effector configured to controllably grasp or release the object; and, a platform configured to support the object.

[0030] In one embodiment: the deployment apparatus includes at least one of: a robot arm including an end effector configured to controllably grasp or release the object; and, a deployment platform configured to present the object to the receiver apparatus; and, the receiver apparatus includes at least one of: a robot arm including an end effector configured to selectively hold the object; and, a receiver platform configured to receive the object from the deployment apparatus.

[0031] In one embodiment the system includes a docking arrangement configured to allow the first and second vehicles to dock prior to transferring the object.

[0032] In one embodiment the one or more processing devices are configured to control the vehicles during load transfer to maintain the set relative pose of the first and second vehicles.

[0033] In one embodiment the one or more processing devices are configured to control at least one of the vehicles and the object transfer mechanism during load transfer to maintain at least one of: an orientation of the object during the transfer; and, a balance of the system during transfer.

[0034] In one embodiment the one or more processing devices are configured to: control at least one of the vehicles to approximately position the vehicles in relative proximity; and, use the pose detection system to determine a relative pose of the vehicles once the vehicles are in relative proximity.

[0035] In one embodiment the one or more processing devices are configured to: determine a second position of the second vehicle; and, navigate the first vehicle to a first position in relative proximity to the second position.

[0036] In one embodiment the one or more processing devices are configured to determine a position of the first and second vehicles using at least one of: inertial sensors; and, position sensors.

[0037] In one embodiment the one or more processing devices interact with a flight controller of at least one of the vehicles.

[0038] In one broad form, an aspect of the present invention seeks to provide an object transfer method for transferring an object between autonomous aerial vehicles, the method including, in one or more processing devices: using pose detection system configured to detect a relative pose of the vehicles to determine a relative pose of the vehicles; controlling at least one of the vehicles in accordance with the determined relative pose so that the vehicles have a target relative pose; and, operating an object transfer mechanism configured to transfer an object between the vehicles to thereby transfer the object between the vehicles.

[0039] It will be appreciated that the broad forms of the invention and their respective features can be used in conjunction and / or independently, and reference to separate broad forms is not intended to be limiting. Furthermore, it will be appreciated that features of the method can beperformed using the system or apparatus and that features of the system or apparatus can be implemented using the method.Brief Description of the Drawings

[0040] Various examples and embodiments of the present invention will now be described with reference to the accompanying drawings, in which: -

[0041] Figure 1 is a schematic side view of an example of an object transfer system configured to transfer an object between autonomous aerial vehicles;

[0042] Figure 2 is a flow chart of an example of a process for transferring an object;

[0043] Figure 3A is a schematic diagram of an example of a processing system for use in the object transfer system of Figure 1;

[0044] Figure 3B is a schematic diagram of an example of a flight controller for a vehicle used in the object transfer system of Figure 1;

[0045] Figure 4A is a schematic side view of an example of a first vehicle including a deployment apparatus;

[0046] Figure 4B is a schematic side view of an example of a second vehicle including a receiver apparatus;

[0047] Figure 4C is a schematic side view of the first vehicle of Figure 4A and second vehicle of Figure 4B when transferring an object;

[0048] Figures 5A and 5B are a flow chart of an example of a process for transferring an object using the vehicles of Figures 4A and 4B;

[0049] Figures 6A to 6C are schematic diagrams of an example of object transfer using a first alternative object transfer mechanism;

[0050] Figures 7A to 7C are schematic diagrams of an example of object transfer using a second alternative object transfer mechanism;

[0051] Figures 8A to 8C are schematic diagrams of an example of object transfer using a third alternative object transfer mechanism;

[0052] Figures 9A to 9D are schematic diagrams of an example of object transfer using a fourth alternative object transfer mechanism;

[0053] Figures 10A to 10D are schematic diagrams of an example of object transfer using a fifth alternative object transfer mechanism;

[0054] Figure 11 is a schematic diagram of an example of an operational workflow of an EBS drone;

[0055] Figure 12A is an image of a specific example of a battery transfer mechanism in use;

[0056] Figure 12B is a first image of the EBS drone of Figure 12A;

[0057] Figure 12C is a second image of the EBS drone of Figure 12A;

[0058] Figure 13 is a schematic example of a control mechanism for the EBS drone of Figure 12A;

[0059] Figure 14A is a first image of a specific example of the receiver drone of Figure 12A;

[0060] Figure 14B is a second image of the receiver drone of Figure 12A with the chute deployed;

[0061] Figure 15 is a flow chart of an example of a control process for transferring an object;

[0062] Figure 16 is a schematic diagram of an example of receiver drone tracker pixel coordinate system;

[0063] Figure 17 A is a schematic plan view of an example of a receiver drone including visual markers;

[0064] Figure 17B is a schematic close up view of an example of a visual marker;

[0065] Figure 17C is a schematic plan view of an example of a visual marker configuration;

[0066] Figures 18A and 18B are images illustrating the detection of drone coordinates and distances using a vision coordinate system;

[0067] Figure 19 is an example of a localisation algorithm;

[0068] Figures 20A to 20D are schematic diagrams of an indoor GPS coordinate system to detect drone coordinates and distances;

[0069] Figure 21 A is a schematic diagram of an example of a drone displacement experiment;

[0070] Figures 2 IB to 2 ID are images captured during the drone displacement experiment of Figure 21 A;

[0071] Figure 22 is a graph illustrating results of a drone displacement experiment;

[0072] Figures 23A and 23B are graphs illustrating results of further drone displacement experiments;

[0073] Figures 24A and 24B are examples of images of marker position estimates for receiver drones;

[0074] Figure 25 is a graph illustrating results of orientation angle correction measurements;

[0075] Figures 26A to 26C are images of an example of object transfer between two drones captured a first test flight;

[0076] Figures 26D to 26F are images of an example of object transfer between two drones captured a second test flight;

[0077] Figure 27 is a graph illustrating results of position correction measurements; and,

[0078] Figures 28A and 28B are graphs illustrating results of vibration measurements during object transfer.Detailed Description of the Preferred Embodiments

[0079] An example of an object transfer system will now be described with reference to Figures 1 and 2.

[0080] The system includes at least two UAVs or other similar vehicles 100A, 100B, with two being shown in this example of the purpose of illustration only. The exact form of the vehicles will vary depending on a range of factors, such as the nature of the pay load and the distance over which the payload is being carried, but typically the vehicles are multi-rotor drones or similar.

[0081] An object transfer mechanism 110 is provided, which is configured to transfer an object O between the vehicles. The object transfer mechanism 110 can be of any suitable form depending on the preferred implementation, but in one example, the mechanism 110 includes a deployment apparatus 111 mounted on a first one of the vehicles and a receiver apparatus 112 mounted on a second one of the vehicles, allowing the object to be transferred from the deployment apparatus 111 to the receiver apparatus 112. The nature of the deployment and receiver apparatus 111, 112 will vary depending on the preferred implementation and specific examples will be described in more detail below.

[0082] The system further includes a pose detection system 115 configured to detect a relative pose of the vehicles. The nature of the pose detection system 115 will vary depending on the preferred implementation, and could include components provided on either one or both of the vehicles 100A, 100B. For example, the pose detection system 115 could include a sensor mounted on one of the vehicles 100A, 100B, which is configured to capture data that can be used to determine the pose of the other vehicle. Typically the pose detection system 115 includes an imaging device, such as a camera, mounted on one vehicle 100 A, 100B, which is configured to capture images of the other vehicle 100B, 100A, including specific information displayed thereon, allowing the relative pose of the vehicles to be calculated, as will be described in more detail below.

[0083] The system typically includes one or more processing devices, which may form part of one or more processing systems, or the like. In one example, this can incorporate flight control systems of the vehicles, but additionally and / or alternatively include one or more separate system controllers, which could be mounted onboard and / or be separate from the vehicles, as will be described in more detail below. Whilst the system can use multiple processing devices, with processing performed by one or more of the devices, it will be appreciated that alternatively a single processing device or system could be used. For the purpose of ease ofillustration the following examples will refer generally to processing devices, but it will be appreciated that reference to a singular processing device should be understood to encompass multiple processing devices and vice versa, with processing being distributed between the devices as appropriate.

[0084] In use, the processing devices are configured to control the object transfer process, and an example of this will now be described with reference to Figure 2.

[0085] In this example, at step 200, the processing devices use the pose detection system 115 to determine a relative pose of the vehicles. For example, this could include analysing images captured by an imaging device, and using results of the analysis to determine the pose of the other vehicle relative to the imaging device. A known configuration of the imaging device can then be used in order to allow a relative pose of the vehicles to be established.

[0086] At step 210, the processing devices control at least one of the vehicles in accordance with the determined relative pose so that the vehicles have a target relative pose. Specifically, this will typically involve controlling one or both of the vehicles to adjust their relative pose, measuring the new relative pose and repeating this process until the target relative pose is achieved. The particular configuration of the target relative pose will vary depending on the preferred implementation and in particular will vary depending on the nature of the object transfer mechanism and the pose detection system. Typically however the target pose includes providing one of the vehicles at least partially above and laterally offset from the other vehicle. This configuration can provide a number of benefits, including facilitating pose detection and object transfer, reducing the impact of downwash from one vehicle on the other, and also mitigating risks associated with vehicle collisions.

[0087] At step 220, once the target pose has been acquired, the processing devices operate the object transfer mechanism to thereby transfer the object between the vehicles.

[0088] Accordingly, it will be appreciated that the above described arrangement enables object transfer between autonomous aerial vehicles, and in particular multi-rotor drones, in flight. This can be used for a wide variety of purposes, such as transferring an object between vehicles allowing these to transport an object in a relay fashion, or to allow transfer of batteries, enabling the flight time of vehicles to be extended.

[0089] In the above arrangement, the use of the pose detection system 115 to detect a relative pose of the vehicles can be used to accurately position the vehicles prior to the object transfer being performed. This can facilitate the transfer and reduce interference between the vehicles, increasing the likelihood of a successful transfer procedure.

[0090] It will be appreciated that in the above example, transfer is described as a one way process from a first vehicle 100A to a second vehicle 100B. However, this is not essential, and it will be appreciated that suitable configuration could allow for bi-directional transfer. For example, this could be achieved through appropriate configuration of the deployment and receiver apparatus, or by having both a deployment and receiver apparatus on each vehicle. Such bi-directional transport could be used, for example, to allow a first vehicle 100A to supply a replacement battery to a second vehicle 100B, and then collect a depleted battery from the second vehicle 100B, allowing the depleted battery to be collected for charging and to reduce the in-flight weight of the second vehicle 100B.

[0091] A number of further features will now be described.

[0092] A specific example of the one or more processing devices configured as part of a control system will now be described with reference to Figures 3 A and 3B.

[0093] In this example, the control system includes at least one system controller 320, which is configured to analyse signals from the pose detection system 115 and vehicle flight controllers 330, and to control the object transfer mechanism 110 and the vehicles 100A, 100B, via respective flight controllers 330.

[0094] The system controller 320 typically includes at least one processing device, such as a microprocessor 321, connected to a memory 322, an input / output interface 323, and a sensor interface 324. The sensor interface 324 is used for connecting the processing device 321 to the pose detection system 115, and optionally any other sensors, such as altitude sensors, position sensors, or the like.

[0095] The input / output interface 323 can be used to receive commands from an external system, such as a radio control unit, computer system, or the like. This enables external commands to be provided to the system, for example allowing the system to be controlled forexample to allow an operator to provide an indication that transfer of the object can commence, provide information regarding a target second vehicle intended to the receive the object, or the like. The input / output interface 323 can also be used to interconnect multiple system controllers 320, to connect the system controller 320 to one or more of the flight controllers 330. Although a single interface is described, in practice functionality may be provided by a number of distinct input / output interfaces 323, such as a serial interface, a network interface, wireless interface, or the like.

[0096] In use, the processing device 321 executes applications software stored in the memory 322, allowing the processing device 321 to perform necessary functions, including receiving data from the pose detection system 115, processing and interpreting the data, and generating instructions for the flight controllers 330 or object transfer mechanism 110.

[0097] The flight controllers 330 typically include at least one processing device, such as a microprocessor 331, connected to a memory 332, an input / output interface 333, a sensor interface 334, and a drive system 338, which is used to control the vehicle.

[0098] The sensor interface 334 is used for connecting the processing device 331 to one or more on-board sensors. The sensors could include an altitude sensor 335, a positional sensor 336 such as a Global Positioning System (GPS) Sensor, and a motion sensor 337, such as an Inertial Measurement Unit (IMU). However, it will be appreciated that a range of other sensors could be provided, and the example arrangement is for the purpose of illustration only.

[0099] The nature of the drive system 338 will depend on the particular vehicle, but as mentioned above, this typically includes a number of rotor motors, for example in the case of a multi-rotor drone. The nature of the drive system is not important for the purposes of the current explanation and it will be appreciated that a wide range of different arrangements could be employed depending on the preferred implementation.

[0100] The input / output interface 333 can be used to communication with the system controller 320, and / or could be used to provide connectivity to other external controllers, computer systems, or the like. This enables commands to be provided to the vehicle flight controller 330 to allow the vehicle to be controlled. For example, this could include providing general commands, such as instructing the vehicle to fly along a particular heading, or to a particularlocation or set of coordinates, with the processing device 331 then generating the necessary control commands to control the drive system 338 as required.

[0101] In use, the processing device 331 executes applications software stored in the memory 332, allowing the processing device 331 to perform necessary functions, including receiving data from the sensors 335, 336, 337, processing and interpreting the data, and making any required decisions. The processing device can also receive input commands from the system controller 320 and generate control instructions for the drive system 338.

[0102] Additionally, processing could be distributed between the processing devices of different vehicles as needed. For example, system controllers 320 could be provided for each vehicle, so that each vehicle 100A, 100B includes a respective system controller 320that controls that the flight controller 330 of that vehicle. Alternatively a single system controller 320 could be used to control multiple vehicles, for example, by interfacing with flight controllers 330 of both vehicles 100A, 100B, or by having the vehicles configured to in a master slave configuration with the system controller 320 controlling a first one the vehicles 100A by interfacing directly with the flight controller 330 of the first vehicle 100A, whilst the other vehicle 100B is slaved to the first vehicle 100A, so that the flight controller 330 of the second vehicle 100B receives commands from the first flight controller 330. This can assist with scaling the system making it easier to employ larger numbers of vehicles.

[0103] In one example, the pose detection system 115 includes an imaging device mounted on one of the vehicles and a visual marker mounted on the other one of the vehicles. For example, the imaging device could be mounted on a first vehicle 100A, which deploys the object, with one or more visual markers being provided on the second vehicle 100B, which receives the object. However, this is not essential, and a reversed configuration could be used with the imaging device mounted on the second vehicle 100B and the visual marker provided on the first vehicle 100A.

[0104] In this example, the processing devices are configured to determine a relative pose of the vehicles by using the imaging device to capture images of the visual marker and then analysing the images to determine a relative pose of the first and second vehicles. Specifically, the images can be analysed to detect the appearance of the marker in the image and using this to calculate the pose of the second vehicle relative to the imaging device. In this regard, if themarker has a known appearance, such as a known size and shape, the size and shape of the marker in the image can be used to derive a geometric transformation indicative of the relative pose of the imaging device and second vehicle. Once this has been determined, information regarding the pose of the imaging device relative to the first vehicle being used to calculate the relative vehicle pose.

[0105] The imaging device may be movably mounted to the first vehicle, for example using a gimbal mounting or similar, which can be used to facilitate imaging of the visual markers, for example allowing a field of view of the imaging device to be adjusted so as to track the visual markers as the two vehicles move relative to each other. In this case, the processing devices also use information regarding movement of the imaging device relative to thereby determine the relative orientation of the vehicles.

[0106] The use of visual markers can assist in more accurately identifying the pose of the second vehicle than can be achieved using other mechanisms. For example, attempting to identify the pose of the second vehicle based on images of the vehicle alone can be problematic as different vehicles might have different appearances in different ambient conditions, for example depending on the orientation of incident light.

[0107] As mentioned above, the processing devices can determine a relative pose of the vehicles based on a geometry of the visual marker in the captured images, and multiple different markers can be used, to help increase the accuracy of the detection process.

[0108] In one example, the visual markers can include machine readable coded data, which can further assist with this process. For example, the visual markers can include coded data indicative of a visual marker identity, with the one or more processing devices using this to determine the visual marker identity, and hence the visual marker size and shape.

[0109] In one particular example, the visual marker includes multiple visual markers and the processing devices being configured to determine the relative pose at least in part based one or more of relative sizes of the multiple visual markers, relative positions of the multiple visual markers, relative orientations of the multiple visual markers and / or identities of the multiple visual markers. The use of multiple markers in this fashion can increase the accuracy of the pose detection process, specifically allowing markers at fixed relative, but physically separatelocations, on the second vehicle to be detected. This increases the likelihood of successful detection of the markers and also improves the accuracy of the pose detection by extending the physical extent over which the detection is performed.

[0110] As previously mentioned, the object transfer mechanism typically includes a deployment apparatus mounted on a first vehicle 100A and a receiver apparatus mounted on a second one of the vehicles 100B. The deployment and receiver apparatus could be of any appropriate form, and could include robot arms, including an end effector to grasp the object. Typically however such systems are complicated and require a high degree of accuracy and control in order to transfer the object, particular as the vehicles might be subject to external forces, such as wind loading, or the like, which can in turn lead to unwanted relative movement of the vehicles.

[0111] Accordingly, in one example, the object deployment apparatus includes a deployment member, such as a slide, configured to transfer the object to the receiver apparatus, for example by allowing the object to slide along the deployment member to the receiver apparatus. The use of a slide is particularly beneficial as this can be used to effect rapid transfer of the object, with the speed of transfer being controlled based on an angle of the slide and level of friction between the slide and object. This minimises the time the vehicles are required to be maintained in the target relative pose. Additionally, the slide does not require a high degree of accurate positioning in order to operate correctly, allowing this arrangement to accommodate minor movements away from the target relative pose.

[0112] Furthermore, the deployment member can be configured to selectively engage the object receiver apparatus, for example by having the deployment member magnetically engage the object receiver apparatus, which helps ensure the deployment member is held stationary relative to the receiver apparatus even in the event of minor relative movements of the vehicles.

[0113] In one example, the deployment apparatus includes a deployment actuator configured to move the deployment member between extended and retracted positions, allowing the deployment member to remain retracted when not use.

[0114] The object transfer system can include a housing configured to retain and optionally secure the object until transfer of the object is to proceed. A release actuator can then beprovided which is configured to release the object from the housing in response to commands from the processing systems. In one particular example, the housing can include multiple slots, each being configured to retain a respective object, with the processing devices being configured to control respective release actuators to thereby release the object from one of the slots. This can be used for example, to allow a first vehicle to carry multiple spare batteries so that it can deploy spare batteries to multiple second vehicles.

[0115] The object receiver apparatus can also include a receiver member, such as a chute, configured to receive the object from the deployment apparatus. In one example, this is configured to engage the deployment member, so that the object can slide directly from the deployment apparatus to the receive apparatus. The receiver apparatus can also include a receiver actuator configured to move the receiver member between extended and retracted positions.

[0116] Using the above described deployment and receiver apparatus arrangements, the processing devices can be configured to determine once the first and second vehicles have the target relative pose and then operate the deployment and receiver actuators so that the deployment and receiver members extend and engage. Once this docking process has been completed, the processing devices then operate a release actuator to release the object from a housing to thereby transfer the object from the first vehicle to the second vehicle.

[0117] As mentioned above, the target relative pose is typically arranged so that a first vehicle is positioned above and laterally offset from a second vehicle to thereby minimise an impact of downwash from the first vehicle. In this example, the pose detection system 115 typically includes a camera mounted on the first vehicle, which is angled downward and outwardly from first vehicle, so that it can image visual markers provided on an upper surface of the second vehicle. The slide and chute of the deployment and receiver apparatus are then orientated so that these align when the vehicles are in the target relative pose, allowing these to provide a substantially continuous surface for the object to traverse.

[0118] In another example, the object transfer mechanism includes a robot arm including an end effector configured to controllably grasp or release the object and / or a platform configured to support the object. Specifically, the deployment apparatus could include a robot arm or a deployment platform, whilst the receiver apparatus could also include a robot arm or a receiverplatform configured to receive the object from the deployment apparatus. This allows for a variety of different arrangements, such as exchanging an object between two robot arms, or using a robot arm to place an object on, or retrieve an object from, a platform.

[0119] The use of robot arms in this manner has a number of potential benefits. For example, the use of robot arms provides greater flexibility in terms of the relative pose of the vehicles during the transfer, as pose variations can be accommodated through suitable robot arm kinematics. This can also facilitate more straightforward exchange of objects, for example allow platform to act to receive an object and also present an object for deployment. However, the robot arms are typically more complex than the slide mechanism previously described. Additionally, the robot arm arrangements may have reduced tolerance for relative movement of the vehicles during the object transfer. Specifically, this can require that the relative pose of the vehicles is held in a fixed position for the transfer process, whereas the slide can accommodate some relative movement of the vehicles.

[0120] In one example, the processing devices are configured to control the vehicles during load transfer to maintain the set relative pose of the first and second vehicles, which can be achieved through ongoing pose detection and / or by employing a leader follower arrangement, so that movement of the second vehicle tracks movement of the first vehicle. Additionally, and / or alternatively, the vehicles can include a docking arrangement configured to allow the first and second vehicles to dock prior to transferring the object. In this example, the vehicles are mechanically held in a fixed relative position during the object transfer to thereby prevent relative vehicle movement. It will be appreciated that a separate docking arrangement could also be used for the slide mechanism described above, although this is generally less important as the slide mechanism can employ engagement between the slide and chute to provide a similar function.

[0121] In one example, the one or more processing devices can be configured to control at least one of the vehicles and the object transfer mechanism during load transfer to maintain an orientation of the object and / or to maintain a balance of the system during the transfer. This can be important to take into account the change in weight distribution as the load moves from one vehicle to the other, whilst maintaining the orientation can be important for some objects, such as those containing a liquid, or the like. Maintaining the orientation and / or balance canbe achieved through appropriate control of both the vehicles and the transfer mechanism. For example, when using robot arms, the robot arm kinematics can be programmed to maintain a constant orientation of the object during transfer, whilst the lift of each vehicle is adjusted to compensate as the weight distribution changes. As part of this process, the processing devices may use signals from inertial sensors, weight distribution sensors, or the like, to detect changes in the vehicle poses, controlling the vehicles as needed to counteract such changes and maintain the object orientation, and / or overall system balance.

[0122] In one example, the processing devices control at least one of the vehicles to approximately position the vehicles in relative proximity and then use the pose detection system 115 to determine a relative pose of the vehicles once the vehicles are in relative proximity. For example, this can be achieved by having the processing devices determine a position of the second vehicle (hereafter a "second position"), for example using on-board position and / or inertial sensors, and then navigate the first vehicle to a position (hereafter a "first position") proximate to the second position.

[0123] A specific example of an object transfer system will now be described with reference to Figures 4 A to 4C.

[0124] In this example, the arrangement includes first and second vehicles 400A, 400B, shown in Figures 4A and 4B, respectively.

[0125] The first vehicle includes an object deployment system 411 including a housing 411.1, and a slide 411.2. In this example, the slide 411.2 is shown extending from the housing 411.1, although it will be appreciated that this could be retractable depending on the preferred implementation. The housing 411.1 is hingably mounted to the vehicle 400A, with an actuator 411.3, such as rotary or linear actuator, being provided to allow the housing to rotate relative to the vehicle.

[0126] The second vehicle 400B includes an object receiving system 412 including a platform 412.1 and a chute 412.2. In this example, the chute 412.2 is shown folded back over onto the platform 412.1 with an actuator 412.3, such as rotary or linear actuator, being provided to allow the chute to be extended outwardly from the platform 412.1. The platform might also be rotatably mounted to the vehicle 400B, allowing this to align with the chute.

[0127] In use, when the vehicles are in the target relative pose as shown in Figure 4C, the slide 411.2 and chute 412.2 can be magnetically engaged via magnetic material located in ends of the slide and chute, allowing an object O to slide from the housing 411.1 onto the platform 412.1.

[0128] Pose detection is performed using a camera 415 mounted on the first vehicle 400A, which is orientated to capture images of visual markers provided on an upper surface of the platform and / or on an underside of the chute, when the chute is folded back over onto the platform.

[0129] An example object transfer process will now be described in more detail with respect to Figures 5 A and 5B. For the purpose of illustration, in this example, it is assumed that the second vehicle 400B is in flight and the first vehicle 400A is used to provide an object, such as a spare battery, to the second vehicle. It is further assumed that the first vehicle 400A includes a first system controller 320A that interfaces with a first flight controller 33OA, whilst the second vehicle 400B includes a second system controller 320B that interfaces with a second flight controller 33 OB.

[0130] At step 500, the first system controller 320A determines the position of the second vehicle, for example receiving this information based on user input and / or from second system controller 320B. At step 505, the first vehicle 400A is controlled to navigate to a position near the second vehicle 400B, with the first system controller 320A determining if the first vehicle 400A is in proximity to the second vehicle 400B at step 510. If not control of the first vehicle continues until general proximity is reached.

[0131] Once proximity has been achieved, at step 515 the first system controller 320A acquires images from the camera 415, analysing these to determine if the second vehicle and then visual markers can be identified within the images at step 520. If not, the first system controller 320A controls the first vehicle 400A, adjusting its position until images of the markers are acquired.

[0132] Once images of the markers are acquired, at step 525 the first system controller 320A calculates the relative poses of the first and second vehicles. For example, this process will typically involve having the first system controller 320A decode the markers to determine their identity, ascertain the size and shape of the markers, and then perform analysis of the imagesto calculate a geometric transformation between the images of the markers and the expected marker size, shape and configuration. This is then used to calculate the relative pose of the vehicles.

[0133] At step 530 the first system controller 320A determines if the relative pose corresponds to a target pose, and if not the first vehicle is controlled to adjust the relative poses, with this being repeated until the target pose is reached.

[0134] At step 535, once in the target relative pose, the first and second system controllers 320A, 320B deploy the slide 411.2 and chute 412.2 by controlling the actuators 411.3, 412.3 respectively, so that these slide and chute magnetically engage. Engagement can optionally be confirmed at step 540, for example, using some form of engagement detection, such as a sensor provided on the slide or chute, or by using images captured by the camera 415. At step 545, the first system controller 320A releases the object from the housing 411.1, allowing this to slide onto the platform 412.1. Successful transfer can optionally be confirmed, for example using a sensor provided on the platform 412.1, and / or by using images captured by the camera 415.

[0135] The first and second system controllers 320A, 320B then retract the slide 411.2 and chute 412.2 by controlling the actuators 411.3, 412.3 at step 555, before the vehicles are controlled allowing them to continue flight as needed at step 460, for example allowing the second vehicle to continue operations and / or allowing the first vehicle to deliver further objects and / or return to base.

[0136] Further example transfer mechanisms will now be described with reference to Figures 6 to 10.

[0137] In the example of Figures 6A to 6C, two vehicles 600A, 600B are provided, each including a robot arm 611, 612, having an end effector 611.1, 612.1. In this example, the first vehicle 600A transports the object O to the second vehicle 600B, either carrying this using the end effector 611.1, or otherwise carrying this and then grasping the object with the end effector to initiate the transfer, as shown in Figure 6A. The first and second vehicles 600A, 600B then move into the target relative pose allowing the second robot arm 612 to be moved so the second end effector 612.2 can grasp the object as shown in Figure 6B. The first end effector 611.1can then release the object, allowing the second vehicle 600B to depart with the object. It will be appreciated that a reverse process can be performed to allow the object to be transferred from the second vehicle to the first vehicle.

[0138] In the example of Figures 7A to 7C, two vehicles 700A, 700B are provided, with a first vehicle 700A including a robot arm 711 having an end effector 711.1, whilst the second vehicle 700B includes a platform 712. In this example, the object O is deployed from the platform 712 and received by the first vehicle 700A, as shown in Figure 7 A. The first and second vehicles 700A, 700B move into the target relative pose allowing the robot arm 711 to be moved so the end effector 711.1 can grasp the object as shown in Figure 7B. The first vehicle 700A can then depart with the object. It will be appreciated that a reverse process can be performed to allow the object to be transferred from the first vehicle to the second vehicle, as shown in Figures 8A to 8C.

[0139] In the example of Figures 9A to 9D, two vehicles 900A, 900B are provided, each including a robot arm 911, 912, having an end effector, and also including a docking mechanism 916A, 916B. The nature of the docking mechanism will vary depending on the preferred implementation, but in one example, the mechanism includes docking plates extending from the vehicles that magnetically engage.

[0140] In this example, the first vehicle 900A transports the object to the second vehicle 900B. The first and second vehicles then 900A, 900B then move into the target relative pose at which point the docking mechanism engages, as shown in Figure 9B. Once this has been completed the second robot arm 912 is moved so that the second end effector can grasp the object as shown in Figure 9C. The first end effector can then release the object, the docking mechanism disengaged, allowing the second vehicle 900B to depart with the object as shown in Figure 9D. It will be appreciated that a reverse process can be performed to allow the object to be transferred from the second vehicle to the first vehicle.

[0141] In the example of Figures 10A to 10D, two vehicles 1000A, 1000B are provided, with a first vehicle 1000A including a robot arm 1011 having an end effector 1011.1, whilst the second vehicle 1000B includes a platform 1012. In this example, the first vehicle 1000A transports a first object Oi, such as a replacement battery, for delivery to the second vehicle 1000B, whilst a second object O2, such as a used battery, is deployed from the platform 1012and received by the first vehicle 1000A, as shown in Figure 1OA. The first and second vehicles 1OOOA, 1OOOB move into the target relative pose allowing the robot arm 1011 to be moved so as to position the first object Oi on the platform 1012, as shown in Figure 10B. The end effector can release the first object, before the first vehicle 1000A is moved to a second target pose, allowing the robot arm 1011 to be positioned so the end effector 1011.1 can grasp the second object Oi as shown in Figure 10C. The first vehicle 1000A can then depart with the second object O2, thereby allowing a battery swap or other similar exchange to be performed as shown in Figure 10D.

[0142] A specific example implementation and associated testing will now be described in more detail.

[0143] For the purpose of this example, an EBS (Emergency Battery Service) drone is developed that is capable of carrying multiple batteries and transferring a battery to an exhausted drone. The drone employs a battery transfer mechanism and drone localization that uses a Cross Marker Position (CMP) method. This arrangement can be used to provide an efficient, balanced transfer that precisely localizes the receiver drone and manages airflow effects due to the proximity between the drones.

[0144] Specifically, in one example, the arrangement employs a diagonal battery swap model that mitigates the airflow turbulence caused by the drones by reducing the distance to 0.5 m between two drones.

[0145] Additionally, the CMP localization method can provide a 97 percent detection rate and enables relative pose lock for the EBS drone. In one version of this, a 70 mm marker is detected up to a 3m distance with a maximum camera tilt angle of 45 degrees. The performance of the transfer mechanism is validated experimentally by successful mid-air transfer in 5 seconds, where the EBS drone is within 2 m distance from the receiver drone, and wherein the resulting 2m / s turbulence does not affect the battery transfer process.

[0146] In this example, the proposed system consists of a helper EBS drone carrying multiple replacement batteries capable of transferring these to a swarm of drones performing aerial sensing applications, thus enabling continuous operation. Figure 11 illustrates the operational workflow of the novel Emergency Battery Service (EBS) technique presented herein.

[0147] Mid-air drone battery swaps offer several advantages for drone swarms. First, they enable continuous operation, allowing drones to stay in the air without landing and recharging, maximizing overall swarm efficiency. Second, battery swaps minimize downtime, ensuring a continuous flow of active drones and maintaining swarm density and performance. Third, by exchanging batteries, drones can cover considerable distances and operate in areas beyond their single-flight range, expanding their mission range. Fourth, mid-air battery swaps provide flexibility for adapting to changing mission requirements, allowing drones to be recharged or equipped with specialized batteries or payloads. Finally, battery swaps streamline swarm scalability, facilitating the addition or removal of drones without disrupting overall swarm operations.

[0148] The described arrangements allow for mid-air drone battery handoff from a single drone to a swarm of drones. This work requires drone localization and docking, aerodynamic analysis, and inter-drone handoff mechanism.

[0149] Drone localization and docking utilises a rough global position estimation is necessary to bring the vehicles close, while a precise relative position determination is required for the actual docking process. There are several methods of drone localization and docking for aerial manipulation used in the literature, where GPS is the most commonly used. Other systems used sensor fused data from inertial sensors along with a computer vision model, as described for example in Jain, K.P. et al. 2020, "Docking two multirotors in midair using relative vision measurements", (Nov. 2020).

[0150] Computational fluid dynamics (CFD) simulations have been used to analyse flow patterns and forces present during mid-air battery transfer. CFD for drones has been studied for applications such as agriculture and vertical measurement accuracies. These simulations offer a comprehensive analysis of the airflow dynamics, encompassing critical factors such as velocity profiles, pressure distributions, and the formation of vortices resulting from the drone interaction. Through these simulations, researchers can explore the intricate details of how the drones' movements and positions influence the airflow dynamics and ultimately impact the success of the battery transfer process. Moreover, studies have investigated the impact of environmental factors, such as wind speed and direction, on the mid-air battery transfer operation. However, research in aerodynamic analysis for mid-air battery transfer has focusedon understanding the complex airflow interactions between two drones flying in close proximity, precisely when one drone is positioned above and the other below. This configuration introduces significant aerodynamic challenges due to the downward force exerted on the lower drone and the downwash generated by the upper drone.

[0151] In the current arrangement an angular approach is used to mitigate the effects of direct downwash from the drone positioned above. Instead of directly aligning with the upper drone, which can result in a substantial downwash and push the hovering drone away, the drones are offset vertically and horizontally, at a specific angle. This approach ensures better stability and minimizes the disruptive forces caused by downwash, enabling successful mid-air battery transfer between the drones.

[0152] The current approach uses visual-inertial odometry in the form of a CMP method, to ensure precise alignment between two drones during mid-air transfers. Through analysis and experimentation, an optimal distance, angle, and timing required for successful battery transfers under various conditions is considered.SYSTEM DESIGN

[0153] An Emergency Battery service (EBS) and receiver drone architecture will now be described. There are various possible applications for drone system design. Table 1 below provides a reference to configurations of the drone developed for testing purposes.Table 1 : The drone hardware specifications

[0154] This section addresses the unique battery case and transfer slide for the EBS drone (first vehicle) and the receiving CMP mechanism of the drones in a swarm (second vehicles). Thetable highlights that both drones in operation have the same configurations and size. The maximum weight both drones can lift is 4.8 kg.EBS Drone

[0155] Figures 12A to 12C show an EBS drone designed for the proposed application. The EBS drone has an onboard computer (OBC), the NVIDIA Jetson Nano. The OBC controls the flight modes and trajectory of the drone by commanding the flight controller as shown in Figure 13. A housing is provided in the form of a case that can carry three batteries simultaneously, as shown in Figure 12B. The case is equipped with a servo that rotates the case as shown in Figure 12A and dispenses a battery based on an IR sensor which checks which slot is available with the fully charged battery. The EBS drone also has a downward facing depth camera that enables localization.

[0156] The EBS drone includes a 30 of 45 degree deployment slide (slider 1) that is 230mm in length and which attaches to the receiver drone using a magnet. The receiver drone has a chute (slider 2) that is 210 mm in length, and this allows both drones docked to be within a 0.5m distance approximately. It will be appreciated that other slide configurations could be used, such as using slides that are longer / shorter or which have a greater / lesser incline, depending on the preferred implementation and / or intended application. The downward-angled slide allows simplified guided gravity-assisted battery handoff and avoids complex, costly and heavy assemblages such as a robot arm, winch, or flying hot-swap battery if a horizontal handoff was implemented. The battery transfer mechanism allows the transfer within 5 seconds. Furthermore, as discussed in more detail below, the arrangement offsets the drones vertically and horizontally, which obviates issues associated with downwash.

[0157] Figure 13 shows the control mechanism of the EBS drone. It illustrates the main control components, OBC, flight controller, and microcontroller. The OBC ensures the drone's flight mode selection and positioning by communicating with the flight controller using MAVLINK. The flight controller provides flight control and GPS connectivity. The depth camera connects to the OBC, which aids the localization of the receiver drone. The microcontroller Arduino Nano 33 loT controls the battery dispensing case; it communicates with the OBC using a two- way serial communication such as a UART serial communication running the ‘rosserial’ Arduino package.

[0158] Figure 13 also highlights the system procedures for the EBS architecture. The receiver drone initiates the process by sending an emergency signal with accurate GPS coordinates. The EBS system accepts the signal request and transmits the verification signal to confirm communication. With confirmed verification, the receiver drone enables position lock to ensure stability. The EBS drone takes flight to the GPS location and begins marker screening. The drone detects the marker within 3 m range of the receiver drone and commences position lock. It further adjusts the height for the optimum distance for the transfer operation. The EBS drone sends ping data to the receiver drone to allow the synchronization of the transfer slides, and both open the slides simultaneously. The slides align and can be locked using a magnet, the EBS drone then transfers the battery. The slides are closed on completion, the EBS drone can return to the station, and the receiver can continue the intended mission. This structured approach ensures efficient and reliable emergency battery swaps for drones in need.Receiver drone

[0159] The receiver drone is a passive quadcopter that waits in HOVER mode for an EBS drone to provide an extra battery. The receiver drone has the NVIDIA Jetson Nano as the OBC to control the drone flight modes and synchronize with the EBS drone. The top base plate of the drone has the Cross-Marker Position (CMP) design illustrated as shown in Figure 14A. The receiving slide (slider 2) is 210mm, allowing for a 500mm distance between the two drones inclined at 30 or 45 degrees. Again it will be appreciated that other slide configurations could be used, such as using slides that are longer / shorter or which have a grcatcr / lcsscr incline, depending on the preferred implementation and / or intended application.

[0160] The main reason for designing the cross-marker position is to replicate the drone's movement. The EBS drone employs an ‘X’ configuration structure that quickly rotates and adjusts its orientation, aligning with the front, back, left, and right fiducial markers. As seen in Figure 14 A, the CMP design has one marker in the center that is 70x70mm in the center. The central marker allows the EBS to detect the receiver drone from a 3m distance. There are four other markers 30x30mm in size; they provide a reference to the EBS drone to understand the reference position of the receiver. The micro gear motor controls the slide that extends the slide to connect to the EBS drone as shown in Figure 14B. The IR sensor onboard confirms receiving the battery.CMP MODEL

[0161] The cross-marker position localization for the drone position will now be described in more detail. This method uses the receiver drone's 6 Degree of Freedom (DoF) localization using an onboard depth camera.System Overview

[0162] The flowchart shown in Figure 15 outlines the step-by-step process of mid-air battery transfer. It commences with the battery request from the receiver drone and proceeds to transmit relevant GPS coordinates and altitude data to the EBS drone. Upon arrival at the designated location, the EBS drone employs a marker- searching algorithm to locate the marker.

[0163] Based on marker detection success, the drone adjusts its altitude, position, and orientation for precise alignment. A handover signal triggers the opening and unfolding of the docking slides on each drone, facilitating the battery transfer to the receiver drone. Following successful transfer, the docking mechanism closes on each drone, and the EBS drone returns to its home location. The flowchart concludes upon completion of the entire process. The process for the EBS drone to approach the receiver drone in its hover position offers multiple strategies. Initially, the EBS drone leverages knowledge of the receiver drone’s GPS coordinates and altitude to achieve a trajectory that minimizes total turbulence.

[0164] The current operational procedure begins with placing the receiver drone in a predefined hover position with specific GPS coordinates and altitude. Subsequently, this information is relayed to the EBS drone, which conducts an initial coarse grained search for the large marker within a 3-meter range. Upon successful detection, the EBS drone initiates a fine grained correction, achieving a precise alignment within a 0.5-meter range using the small markers in both the horizontal and vertical dimensions for a diagonal handoff. This correction also includes orientation adjustments to ensure the EBS drone is optimally positioned for the item handoff.Visual Inertial Navigation

[0165] Visual odometry estimates the relative camera motion between consecutive frames based on the tracked features. Fusing visual odometry and inertial data gives accurate and robust pose estimates. Visual-Inertial Navigation (VIN) using markers is a technique thatcombines information from visual data (images) and inertial measurements (accelerometer and gyroscope data) to estimate the pose (position and orientation) of a device or camera. Markers are distinctive environmental features or patterns to aid localization and tracking. The first step is detecting and tracking visual features in the camera images, such as corners, edges, or specific patterns. After successfully detecting the markers, the subsequent step is to utilize them to determine the camera’s pose.

[0166] In one example, this process involves accessing the camera’s calibration parameters to achieve camera pose estimation, including the camera matrix and distortion coefficient, with an EBS coordinate system being used to verify the position of the drone. The camera is positioned at 45° on the drone in flight, with the aim being to estimate the pixel coordinate for the receiver drone and depict the deviation in position using the CMP model.

[0167] Three different coordinate systems can be used for this positioning system, namely a receiver-drone coordinate system, an EBS coordinate system, and a frame coordinate system. The drone coordinate system helps determine the positions of points in the flight facility. The coordinate system selects an origin (0, 0, 0) by choosing a corner and defining the X, Y, and Z axes along the ground and vertical dimensions. With this setup, any drone location in the GPS location is in 3D space by measuring its distance from the origin along the X, Y, and Z axes.

[0168] The drone coordinate system is represented in Figure 16. For instance, for the receiver drone R, its coordinates in the drone Coordinate System would be represented as Xr, Yr, Zr. To transform the coordinates of point R from the drone coordinate system to the EBS coordinate system, a rotation matrix R and translation vector t are used. This process expresses the coordinates of point R in the camera’s coordinate system as XEBS, YEBS, ZEBS- Mathematically, the transformation equation is:

[0169] Once the point R is in the camera’ s coordinate system, it can be projected onto the image plane using equations derived from similar triangles. The extrinsic matrix combines the rotation and translation vector to transform the 3D point from the drone coordinate system to the camera coordinate system:E = [7?|t]

[0170] The EBS coordinate system projects the 3D point XEBS, YEBS, ZEBS onto the image plane.This projection results in the image plane’s 2D coordinates x and y.

[0171] The intrinsic matrix (I) represents the process containing the camera’s intrinsic parameters. The camera’s intrinsic parameters, like the focal length (f), are used in the intrinsic matrix to project 3D points onto the image plane, which allows for a more convenient representation and calculation of the image coordinates.

[0172] The camera’s optical center, represented as c_x and c_y, may not coincide with the center of the image coordinate system. This offset indicates that the camera’s principal point (the point where the optical axis intersects the image plane) is not at the center of the image. A slight skew angle a may exist between the x and y axes of the camera sensor, which means that the axes are not perfectly perpendicular to each other, leading to a slight rotation in the image. The intrinsic matrix (I) accounts for these parameters, allowing for an accurate projection and transformation from 3D world coordinates to 2D image coordinates.CMP Localization

[0173] Pose estimation is important for many computer vision applications, including drone navigation. Finding correspondences between areas in the real world and their 2D image projection is the basis of this technique. The position extraction is challenging; using a fiducial marker allows for more accurate positioning.

[0174] A synthetic square marker called an ArUco marker consists of a large black border and an interior binary matrix that establishes its identification (ID). The ArUco markers are available in several dictionaries to determine their ID and size.

[0175] For this application, the markers used are the '4X4 DICT'. Each receiver drone will be equipped with a unique marker design, i.e., the Cross Marker Position (CMP). Figures 17A to17C illustrate the CMP configuration. The CMP method provides reference to five positions in the marker; the top, bottom, left, right and center.

[0176] The CMP method detects the position and ID of all five markers. The EBS drone is equipped with a depth camera that will detect the position and ID of the markers. The camera has been calibrated to resolve any distortion or depth estimation. The position localization is determined using Robot Operating System (ROS). In one example, camera calibration is conducted with standard square checkerboard method. The camera calibration sets the optimum values for the camera matrix, distortion, rectification, and projection. This calibration file is set as the parameters for the detection and localization launch file.

[0177] Figures 18A and 18B shows the detection of markers with position and orientation values.

[0178] The developed package determines the marker frame and the reference position with respect to the camera alignment. The ‘aruco single / pose’ ROS topic will publish the pose and the orientation. The CMP method requires simultaneous detection of five markers and the default reference frame to allow the EBS drone to navigate. The previous section discusses the use of four markers 30mm and one 70mm marker, all derived from of the ‘4X4’ DICT.

[0179] Table 2 shows the syntax for the arguments defined for the launch script, this can be modified based on the application.Table 2: The code syntax for CMP method

[0180] The EBS drone has been set to detect default marker sizes. The ROS ‘tf frame publishes the corresponding marker id. The marker dictionary is also predefined and can be modified.

[0181] An example algorithm for performing CMP localisation is shown in Figure 19.

[0182] The drone positions itself with ‘vel_cmd_y’ velocity commands for moving left andright, and uses ‘vel cmd x’ to move front and back. The ‘yaw rate cmd’ allows the EBS drone to adjust its yaw angle orientation. Angular velocities can be estimated using Euler rates. The equation below shows the correlation between drone angular velocities (p, q, r) with the angular rotation’s values of pitch, roll, and yaw represented as (p, 0, and ip, respectively.

[0183] To estimate the quaternion for drone movements containing four real parameters, they are propagated according to the differential equation below. The matrix form illustrates a product between the angular rates and quaternion values represented as x, y, z, w. These values are extracted using the CMP localization on the receiver drone.

[0184] To calculate the accurate positioning for the EBS drone, the equation below is used. The EBS drone estimates the appropriate orientation and position of the drone using the product of the cosine (c) and sin (s) values of GPS latitude (la) and longitude (lo) values that are represented in the matrix factoring the radius of the earth (R) and the quaternion matrix derived using CMP localization.

[0185] This system can be translated to any drone size as the algorithm can be modified to accommodate the preferred marker size.RESULTS AND VALIDATION

[0186] The section discusses the experiments conducted to perform battery transfer between two drones. The drone position is detected using the developed computer vision model. This model determines the distance between the drones and the optimal position for a balanced transfer. Additionally, the research incorporates marker and camera tilting angle techniques to verify the measured distances between the EBS and receiver drone. The findings highlight the potential of computer vision and innovative drone detection and distance assessment techniques, improving drone navigation and localization.Position Optimization

[0187] The decision about the position of the two drones helps achieve a balanced transfer. While hovering in a fixed position, the multi -rotor generates gales known as "downwash." To investigate the effect of downwash, it was arranged to have one UAV fly beneath another UAV hovering in a fixed place in an experiment shown in Figure 17 A. This confirmed that the downwash from the transport UAV affects the operational UAV underneath it when aerial docking and the experiments validate the destabilization of the drone due to air suction in the drone below. A motion-tracking system using four external beacons was positioned to verify the drone’s location. Figures 20A to 20D shows the three stages of the drone flight wherein the beacons 10 and 11 represent the EBS and Receiver drone, respectively. The table in Figure 20C shows the distance of the drones from each beacon, and the green-coloured Transmission / Receiver mode indicate active communications between all beacons.

[0188] The first stage is the EBS approaching the receiver drone, which is perpendicular to each other for the second stage. The third stage highlights the displacement due to the effects of downwash. These experiments help validate the accurate positioning to execute object transfer.

[0189] The [x,y] pixel coordinates and a distance (height) measurement are the three parameters output in the computer vision system. The effect of downwash is substantially more substantial immediately beneath the UAV and diminishes as one moves away from the center of the UAV. The effect of downwash can be reduced if the UAV beneath it specifies a target point on either side or a specific distance beneath the other UAV. Several flight tests were conducted to test the optimal drone position for battery transfer. Figures 2 IB to 2 ID and Figure22 demonstrate the results of the drone displacement due to the effect of the downwash, with the configuration used in testing being shown in Figure 21 A.

[0190] It can be observed in Figure 2 IB when the drone is positioned at 3m distance between each other, the receiver drone is exposed to the downwash from the EBS drone. Due to the downwash turbulence, the drone below is destabilized and can drift across the x or y axes. The horizontal displacement is captured in Figure 21C and 2 ID.

[0191] An anemometer was used to investigate the wind velocity of downwash under the UAV by varying the height of the bar above the ground from Im to 5m. The altitude measurements were extracted from the flight. These measurements validate the optimal position of the drone.

[0192] The effect of that has been analysed in Figure 22 where the y-axis represents the altitude of the drones in meters and the x-axis represents time in seconds. It is observed that in the hover state flying at an altitude of 2.5m, drone 2 continues to maintain its position-hold state. However, upon drone 1 flying over its state, it starts to experience a downward push when vertically aligned at about 6th second before repositioning to its desired altitude. The optimal position of the receiver drone is observed diagonal to the EBS drone, this allows the EBS drone to be closer. The low distance of 0.5m between two drones allows for a balanced and quick transfer.

[0193] The altitude measurements were extracted from the flight. These measurements validate the optimal position and trajectories of the drone. Figures 23A and 23B show an average displacement of the receiver drone from the EBS drone for different height separations. Here, the y-axis represents the displacement of the drones in meters, and the x-axis represents the distance between drones in meters. In the hover state, at an altitude ranging between 1 to 4 meters, the drone has a very high displacement that gradually decreases as the distance increases.

[0194] 50 iterations of drone flights demonstrate displacement of the drone positioned below (i.e. the Receiver drone) while being approached using four possible trajectories. The effect of that has been analysed in Figure 23B, where points ‘H-V’, ‘H-H’, ‘ V-V’ and ‘ V-H’ represent different trajectories ‘Horizontal-then-Vertical’, ‘Horizontal-Horizontal’, ‘Verticalvertical’, and ‘ Vertical-then-Horizontal’ respectively for the EBS drone to approach the Receiver drone.The scatter plot highlights the displacement values in meters for each trajectory tested for 50 iterations.

[0195] Trajectory ‘H-H’ demonstrates wherein the EBS drone first attempts horizontal alignment of 0.5 m between the two drones and then vertically aligns to 0.5 m ‘alt’, and Trajectory ‘V-H’ demonstrates the proposed trajectory approach wherein the EBS drone first attempts vertical alignment of 0.5m ‘alt’ and then aligns horizontally to 0.5 m distance. However, in ‘H-H’, both drones are moving to align themselves horizontally, assuming the vertical distance of 0.5 m has been achieved. Similarly, in ‘V-V’, the vertical alignment movement has been tested, assuming horizontally, the drones are aligned at a 0.5m distance.

[0196] The trajectories for ‘H-V’ and ‘H-H’ have demonstrated the highest displacement between the drones that are caused due to the downwash, as it leads to several occurrences of more than 50% overlap due to the horizontal approach method, which results in airflow disturbance and change in thrust. However, trajectories ‘V-V’ and ‘V-H’ demonstrate minimal displacement, as the resultant overlap was always 50% or less.

[0197] In real-world testing it is not always possible to predetermine the horizontal alignment, these experiments indicate the ‘Vertical-then-Horizontal’ trajectory to be the most optimal for testing and verification; this allows the EBS drone to be closer. Thus, it is not ideal to transfer between two drones while they are perpendicular to each other due to instability. The low vertical distance of 0.5 m between two drones allows for a balanced and quick transfer.CMP Drone Localization

[0198] In this section, results are presented of experiments aimed at validating the accuracy of the proposed pose estimation system for the EBS drones using CMP localization. The system’s ability to achieve accurate position locking while maintaining the desired orientation relative to the front marker is evaluated.

[0199] Indoor experiments validated the EBS drone positioning of the CMP containing Receiver drone set in different directions. The CMP detection method is illustrated in Figure 24A and 24B, detecting 5 markers with their ID information and the reference position and distance from the camera. The camera output is a default of 30 fps and the time for detection is 1ms.

[0200] The CMP method can detect the 70 mm marker up to 3m in range. The depth camera positioned on the EBS drone modifies the position of the camera and the changes the pose of the marker. The camera was tested for a range of positions to determine the Field of View (FoV) detection range and the optimum tilt angle. The camera can detect the markers within a tilt-range of 30 degrees to 55 degrees. The optimum angle for detection during the battery transfer is observed to be 45 degrees.

[0201] The ground truth positions were obtained through a motion tracking system to serve as a reference for evaluations. The position error for each marker association was calculated to assess the accuracy of position locking. The position error was computed as the Euclidean distance between the drone’s estimated position and the ground truth position of the designated marker. The equation derives the position error ‘e’ between the drone’s estimated and actual ground truth positions. The ‘e’ values are approximately 0.9 cm on an average of 20 iterations, which indicates higher accuracy, implying the drone is effectively locked onto the intended marker position.

[0202] The results demonstrate that the proposed system achieves highly accurate position lock, with errors consistently below a predefined threshold; this indicates that the drone successfully reaches the intended marker positions with minimal deviation. For each marker association e.g., Left- Back, Right - Back, Front - Back, the position error is calculated as follows, where x_est, y_est and z_est are estimated position across all axes and x_IP, y_IP and z_IP are the desired position values:

[0203] The orientation control aspect for the EBS drone was evaluated by analysing the orientation deviation between the drone’s estimated orientation and the desired orientation relative to the front position of the CMP design. This deviation was calculated as the angular difference between the two orientations. Figure 25 presents the orientation deviation analysis results. The orientation deviation measures the angular difference between the drone’s estimated and desired orientation relative to the front marker. It signifies how closely the drone’s heading aligns with the specified orientation. For each marker association, theorientation deviation is calculated as follows, wherein Oest is the estimated drone angle and OiP is the desired position: a = |0_est — 6_IP\

[0204] This calculation yields the deviation value in a, representing how much the drone’s orientation deviates from the desired direction. The a measures < 1 degree over an average of 20 tests indicating that the drone successfully maintains the required position. Figure 25 shows that the system effectively maintains the drone’s orientation within a narrow range of deviation from the desired orientation; this validates the system’s ability to correct its orientation accurately with approximately a degree of error and adhere to the specified orientation constraint.

[0205] The Right-Back plot in Figure 25 highlights orientation correction from the right to the back, the angle measurements vary between 30 to 50 degrees, and the Left-Back plot demonstrates angular correction between -30 to -50 degrees from the left position.

[0206] Table 3 summarizes each marker association’s average position error and orientation deviation to provide a comprehensive system performance overview.Table 3

[0207] The equation derives the position error (e) between the drone’s estimated and actual ground truth positions. The ‘e’ values are approximately 0.9 cm on an average of 20 iterations, which indicates higher accuracy, implying the drone is effectively locked onto the intended marker position. The results demonstrate that the proposed system achieves highly accurate position locking, with position errors consistently below a predefined threshold; this indicates that the drone successfully reaches the intended marker positions with minimal deviation.

[0208] The results in Table 3 reinforce the consistent accuracy of the system across different marker associations. The low average position errors and orientation deviations furtherestablish the system’s capability to achieve accurate position locking while maintaining the correct orientation, validating the CMP design’s effectiveness.Batery Transfer Validation

[0209] In order to validate the effectiveness and feasibility of the proposed battery transfer system, a series of experiments were conducted. The experiments aimed to assess the performance of the visual-inertial odometry method in achieving precise alignment between the drones during mid-air battery transfers. Various test cases were conducted to replicate different environmental conditions and flight dynamics.

[0210] A total of 15 outdoor flight tests or iterations were conducted to replicate different positions and flight dynamics, and images captured during two such test flights are shown in Figures 26A to 26F. In this regard, Figures 26A to 26C shows an EBS drone, docking of the EBS and receiver drone and object transfer, whilst Figures 26D to 26F show the EBS drone, object transfer, and the receiver drone with the received object. Ambient wind conditions were less than 4 m / s. Figure 27 illustrates the position error for various marker associations. The plot line demonstrates the position correction from front to back; the drone varies the position between 4.5 to 15 cm. These measurements account for the distance between the markers in the CMP design to aid the drone’s movement. The position error is approximately 0.9 cm on average, and it measures how accurately the drone’s estimated position aligns with the ground truth position of the designated marker. The drone has demonstrated a successful transfer despite the offset of 0.9 cm.

[0211] The battery transfer mechanism implemented in the EBS drone system enables a smooth and efficient transfer process. With a 30-degree or 45-degree slide on the EBS drone measuring 230mm and a corresponding slide on the receiver drone measuring 210mm, the two drones can securely dock with each other using a magnet. This docking arrangement ensures that the drones remain within an approximate distance of 0.5m during the transfer. The battery transfer itself is completed swiftly, taking only 5 seconds to successfully complete the process. The 5 seconds window consists of opening of the slides, battery handoff and slide closure.

[0212] To estimate the smoothness of the transfer, the amount of vibration encountered during the handoff was measured, with results shown in Figures 28 A and 28B, which show the typicalvibration along the x, y, and z axes experienced during one of the EBS and Receiver drone tests, in the different plots. The vibration has been measured for the transfer time between 3:30 to 3:40. The graphs indicate a stable handover as the vibrations across all axes are close to zero. The experimental data supported the notion that the angular approach, combined with visual- inertial odometry, effectively mitigated the challenges associated with direct downwash and ensured a stable and efficient battery transfer process.

[0213] The system incorporates a sloped transfer mechanism that leverages gravity for efficient battery transfer, sidestepping downdraft instability by diagonally placing the drones. This transfer mechanism is executed at 4m / s windspeed outdoors, and the transfer process was 4 seconds. These validation experiments provide strong evidence for the viability of the proposed system and pave the way for its practical implementation in real-world scenarios.

[0214] The outdoor experiments exposed the system to non-ideal wind conditions of up to 4 m / s. Despite such challenging environmental factors, the vision system navigated through sun glare conditions, ensuring little to no interruptions. Notably, the system exhibited a minimal latency of 1 ms in detection time, underscoring its responsiveness. Furthermore, tests revealed an average localization accuracy of 0.9 cm. These findings collectively show the potential and robustness of our developed system across dynamic conditions.

[0215] Accordingly, the results demonstrated successful battery transfers with high accuracy and reliability, as the drones maintained stable positions and minimized disruptive forces caused by downwash. The experimental data supported the notion that the angular approach, combined with visual-inertial odometry, effectively mitigated the challenges associated with direct downwash and ensured a smooth and efficient battery transfer process. These validation experiments provide strong evidence for the viability of the proposed system and pave the way for its practical implementation in real-world scenarios.

[0216] Accordingly, the above described arrangement provides a technique to address the limitations of drone flight duration by proposing an Emergency Battery Service (EBS) drone capable of transferring batteries to depleted drones in mid-air. The battery transfer mechanism, utilizing a 30-degree slide on the EBS drone and a corresponding slide on the receiver drone, enables secure docking between the two drones using a magnet.

[0217] The experimental results validate the efficiency and reliability of the battery transfer system, with successful transfers completed within 5 seconds and the drones maintaining a distance of approximately 0.5m during the process. The proposed system offers a practical solution to enhance the continuous operation of drone swarms by providing an emergency energy source.

[0218] Furthermore, the approach addresses key challenges associated with drone localization and docking, as well as aerodynamic analysis for mid-air battery transfer. The CMP method, utilizing markers for precise position estimation, enables accurate localization up to 97 percent of the receiver drone, while the angular approach mitigates disruptive forces caused by downwash during the transfer process. Computational fluid dynamics (CFD) simulations are employed to analyse airflow dynamics and optimize the system design. The proposed system's performance is supported by experimental validation, demonstrating the successful mid-air transfer and the ability to maintain stable positions and minimize disruptive forces.

[0219] Overall, this research contributes to the advancement of drone swarm technology by providing a practical solution to overcome the limitations of battery capacity. The proposed EBS drone system with its unique battery transfer mechanism and localization methods offers a reliable and efficient means to extend the flight duration of drones. Further development and implementation of this system could unlock new possibilities for various applications, where continuous drone operation and battery replenishment are essential.

[0220] Throughout this specification and claims which follow, unless the context requires otherwise, the word “comprise”, and variations such as “comprises” or “comprising”, will be understood to imply the inclusion of a stated integer or group of integers or steps but not the exclusion of any other integer or group of integers. As used herein and unless otherwise stated, the term "approximately" means ±20%.

[0221] Persons skilled in the art will appreciate that numerous variations and modifications will become apparent. All such variations and modifications which become apparent to persons skilled in the art, should be considered to fall within the spirit and scope that the invention broadly appearing before described.

Claims

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:1) An object transfer system configured to transfer an object between autonomous aerial vehicles, the system including: a) an object transfer mechanism configured to transfer an object between the vehicles; b) a pose detection system configured to detect a relative pose of the vehicles; and, c) one or more processing devices configured to: i) use the pose detection system to determine a relative pose of the vehicles; ii) control at least one of the vehicles in accordance with the determined relative pose so that the vehicles have a target relative pose; and, iii) operate the object transfer mechanism to thereby transfer the object between the vehicles.2) The object transfer system according to claim 1, wherein the pose detection system includes: a) an imaging device mounted on one of the vehicles; and, b) a visual marker mounted on an other one of the vehicles, and wherein the one or more processing devices are configured to determine a relative pose of the vehicles by: i) using the imaging device to capture images of the visual marker; and, ii) analysing the images to determine a relative pose of the first and second vehicles.3) The object transfer system according to claim 2, wherein the one or more processing devices are configured to determine a relative pose of the vehicles based on a geometry of the visual marker in the captured images.4) The object transfer system according to claim 2 or claim 3, wherein the visual marker includes machine readable coded data and wherein the one or more processing devices are configured to determine the relative pose at least in part using the coded data.5) The object transfer system according to claim 4, wherein the coded data is indicative of a visual marker identity and wherein the one or more processing devices are configured to determine the relative pose at least in part using the visual marker identity.6) The object transfer system according to any one of the claims 2 to 5, wherein the visual marker includes multiple visual markers and wherein the one or more processing devices are configured to determine the relative pose at least in part based on at least one of: a) relative sizes of the multiple visual markers; b) relative positions of the multiple visual markers;c) relative orientations of the multiple visual markers; and, d) identities of the multiple visual markers.7) The object transfer system according to any one of the claims 1 to 6, wherein the object transfer mechanism includes: a) a deployment apparatus mounted on a first one of the vehicles; and, b) a receiver apparatus mounted on a second one of the vehicles.8) The object transfer system according to claim 7, wherein the object deployment apparatus includes a deployment member configured to transfer the object to the receiver apparatus.9) The object transfer system according to claim 8, wherein the deployment member is configured selectively engage the object receiver apparatus.10) The object transfer system according to claim 9, wherein the deployment member is configured to magnetically engage the object receiver apparatus.11)The object transfer system according to any one of the claims 7 to 10, wherein the deployment member includes a slide configured to slidably deploy the object.12) The object transfer system according to any one of the claims 7 to 11, wherein the deployment apparatus includes a deployment actuator configured to move the deployment member between extended and retracted positions.13)The object transfer system according to any one of the claims 7 to 12, wherein the deployment apparatus includes: a) a housing configured to retain the object; and, b) a release actuator configured to release the object from the housing.14)The object transfer system according to claim 13, wherein the housing includes multiple slots, each being configured to retain an object and wherein the one or more processing devices are configured to control the release actuator to thereby release the object from one of the slots.15)The object transfer system according to any one of the claims 6 to 14, wherein the object receiver apparatus includes a receiver member configured to receive the object from the deployment apparatus.16) The object transfer system according to claim 15, wherein the receiver member includes a chute.17)The object transfer system according to claim 15 or claim 16, wherein the receiver apparatus includes a receiver actuator configured to move the receiver member between extended and retracted positions.18)The object transfer system according to claim 7 and claim 15, wherein the one or more processing devices are configured to: a) determine once the first and second vehicles have the target relative pose; b) operate deployment and receiver actuators so that deployment and receiver members extend and engage; and, c) operate a release actuator to release the object from a housing once the deployment and receiver members are engaged to thereby transfer the object from the first vehicle to the second vehicle.19)The object transfer system according to any one of the claims 1 to 18, wherein the target relative pose is configured at least one of: a) so that a first vehicle is positioned at least partially above and laterally offset from a second vehicle; and, b) to minimise an impact of downwash from the first vehicle.20) The object transfer system according to any one of the claims 1 to 7, wherein the object transfer mechanism includes at least one of: a) a robot arm including an end effector configured to controllably grasp or release the object; and, b) a platform configured to support the object.21)The object transfer system according to claim 7 and claim 19, wherein: a) the deployment apparatus includes at least one of: i) a robot arm including an end effector configured to controllably grasp or release the object; and, ii) a deployment platform configured to present the object to the receiver apparatus; and, b) the receiver apparatus includes at least one of: i) a robot arm including an end effector configured to selectively hold the object; and, ii) a receiver platform configured to receive the object from the deployment apparatus.22) The object transfer system according to any one of the claims 1 to 21, wherein the system includes a docking arrangement configured to allow the first and second vehicles to dock prior to transferring the object.23) The object transfer system according to any one of the claims 1 to 22, wherein the one or more processing devices are configured to control the vehicles during load transfer to maintain the set relative pose of the first and second vehicles.24) The object transfer system according to any one of the claims 1 to 23, wherein the one or more processing devices are configured to control at least one of the vehicles and the object transfer mechanism during load transfer to maintain at least one of: a) an orientation of the object during the transfer; and, b) a balance of the system during transfer.25) The object transfer system according to any one of the claims 1 to 24, wherein the one or more processing devices are configured to: a) control at least one of the vehicles to approximately position the vehicles in relative proximity; and, b) use the pose detection system to determine a relative pose of the vehicles once the vehicles are in relative proximity.26) The object transfer system according to claim 25, wherein the one or more processing devices are configured to: a) determine a second position of the second vehicle; and, b) navigate the first vehicle to a first position in relative proximity to the second position.27) The object transfer system according to claim 25 or claim 26, wherein the one or more processing devices are configured to determine a position of the first and second vehicles using at least one of: a) inertial sensors; and, b) position sensors.28) The object transfer system according to any one of the claims 1 to 27, wherein the one or more processing devices interact with a flight controller of at least one of the vehicles.29) An object transfer method for transferring an object between autonomous aerial vehicles, the method including, in one or more processing devices: a) using a pose detection system configured to detect a relative pose of the vehicles to determine a relative pose of the vehicles;b) controlling at least one of the vehicles in accordance with the determined relative pose so that the vehicles have a target relative pose; and, c) operating an object transfer mechanism configured to transfer an object between the vehicles to thereby transfer the object between the vehicles.

Images