Amazon’s Prime Air delivery drones have caused quite a stir, without even entering service. Recently the firm published a proposal for installing docking stations on street lampposts (Figure 1), giving an insight into the kind of logistical challenges its development teams are tackling. As a place for transferring loads, these stations can provide flexibility for delivery planning, and perhaps more importantly, a charging point for the vehicles to re-energize before continuing to a final destination, waiting for a next assignment, or returning to base. The lamppost is also, conveniently, connected to an electrical supply.
Figure 1: Amazon has identified tall lampposts as a safe place for its Prime Air drones to perch.
Precision docking of autonomous vehicles
To recharge, an autonomous vehicle must connect accurately with the charging electrodes provided on the docking station. Amazon’s patent application hints at the type of arrangement it may be considering (Figure 2).
Figure 2: This sketch from the patent application suggests how the drone may connect to the charging point.
To position the vehicle prior to docking, a number of options may be viable. Existing robot-docking systems provide a few examples to consider.
Robotic vacuum cleaners, like the Roomba, have been in the market for some time, and handle battery charging autonomously by monitoring the battery condition during cleaning in order to determine when to return to the base station to recharge. The base station emits an infrared beacon, and the Roomba locks onto this signal keeping the beacon in the center of its field of view as it returns.
The open-source ardumower robotic lawnmower project uses a slightly different technique. When the battery voltage falls below a level indicating charging is necessary, the vehicle will follow a perimeter wire embedded in the ground to return to its charging station. By following the wire accurately, the mower arrives at the entrance to the docking station in a suitable position to be guided towards electrodes inside the dock that make contact with charging pads on the vehicle surface.
Visual alignment is another option that has been demonstrated in various projects. The maker community is a hive of robotic experimentation, including this online video showing how a front-mounted camera is used to locate a charging station and ensure correct alignment. The robot in this demonstration, which uses Raspberry Pi and Arduino computers, stops several times to adjust alignment and is able to correct its position by a few centimeters each time. Fine-tuning the vehicle position may present a more complex challenge when the vehicle is an airborne drone hovering above the docking station. Gusting winds may add further complications.
Wireless charging is an alternative, and is featured in drone landing pads currently marketed by Skysense. Wireless charging is already available for mobile phones and electric vehicles. Charging can proceed if the transmitter and receiver are separated by up to a few centimeters. Closer positioning for faster energy transfer can be achieved by guiding the receiving device to an optimum position or using one of several approaches that allow the power receiver to be positioned freely. These include building multiple coils into the charging pad to generate a magnetic field only in the vicinity of the receiver, or mechanically driving the transmitter coil to a position close to the receiver.
A team at Harbin Institute of Technology in Heilongjiang Province, China, has tackled the challenges in autonomously aligning the modules of a reconfigurable robot system using a combination of visual pre-alignment and subsequent high-accuracy alignment using linear Hall sensors. This robot, called UBot, can reconfigure itself by rearranging modules as necessary to perform various tasks, and is also capable of self-repair by replacing defective modules when needed. A sensor module containing the CCD vision sensor and four linear Hall sensors in positions that correspond to the locations of magnets in the active docking module. The modules are able to move forwards and sideways, and can rotate. The CCD sensor is used for target acquisition and to guide the pre-alignment phase. The signals from the linear Hall sensors are then used to control precise positioning before mechanical mating using a hook mechanism completes the docking procedure.
Closer examination of the Harbin team’s findings gives insights into the challenges intrinsic to visual or magnetic alignment. Visual alignment is accomplished by extracting recognized features from the CCD image, calculating angular and linear offsets, and making linear adjustments until these are within 3 degrees and 4 mm respectively. The active module then moves closer until the captured image occupies 80% of the field of view. Visual alignment then terminates, and precision alignment using the linear Hall sensors commences. X, Y and angular corrections are calculated from the output voltages of the four Hall sensors, in relation to coordinates on the module mating surfaces, to position the modules with accuracy greater than 1 mm in both linear axes and 2 degrees of angular displacement.
A large variety of image sensors or camera modules for machine-vision applications may be able to support visual alignment in robotic applications. The ON Semiconductor NOIV1SN1300A is a monochrome CMOS sensor that integrates a programmable-gain amplifier and 10-bit analog-to-digital converter, and provides four LVDS serial outputs. An alternative variant with parallel CMOS output is also available. The pixel array is 1280 x 1024, and up to eight specific regions of interest can be programmed, allowing high frame rates if needed.
Unlike a digital Hall sensor, the linear type as used in the Harbin project allows displacement to be calculated with high resolution thereby allowing repeated measurement-correction cycles leading ultimately to accurate docking. A linear sensor like the Cherry LIN-11HAW, which is designed for use in industrial drives and controls, as well as automotive gearshift position detectors, has a measurement range of up to 45 mm and dual redundant outputs for enhanced reliability.
Design-out demand for accuracy?
Drawing on existing knowledge for docking and recharging autonomous vehicles, a variety of techniques can be considered to enable a drone to position itself relative to the charging point of a docking station. The typical approach, with wheeled vehicles, has been to mount the sensing system on the robot, using the robot’s own intelligence and propulsion to adjust position relative to the fixed dock. However, the positioning system can add weight to the robot, which may not be acceptable in a drone application. For Amazon, extra weight can translate into reduced carrying capacity and delivery range.
Alignment-free wireless charging may provide a solution that minimizes the need for extra electronics on-board the drone. If ohmic charging contacts are preferred, for example using a waterproof connector, a more sophisticated docking station with inbuilt mechanisms to secure and reposition the drone may provide a workable alternative. This could be as simple as a hook and winch, requiring the drone only to provide a “grab handle” and charging connector in standard positions. This may help to reduce the bill of materials cost for each individual drone, and give freedom to change the drone design in the future if necessary; the trade-off being a more sophisticated docking station that may require regular maintenance.
Conclusion
Any approach to docking an autonomous vehicle for battery charging imposes a requirement for accurate positioning. A number of optical and magnetic position-detection techniques may be considered, although aligning a drone within millimeter accuracy while hovering could prove difficult and may be further complicated by weather conditions such as rapidly varying wind speed. Wireless charging or inventive mechanical engineering may provide a means of designing-out the need for active positioning on-board the drone itself.
