Unmanned Systems Technology 022 | XOcean XO-450 l Radar systems l Space vehicles insight l Small Robot l BMPower FCPS l Prismatic HALE UAV l InterDrone 2018 show report l UpVision l Navigation systems

40 Focus | Radar systems mainstream chip-making processes. It is this baseband processor that delivers the same effect as beam steering, by selecting a ‘beam’ from the large amount of incoming data and handling a 100 Gbit point cloud per image in real time at 30 fps. This ‘digital’ beam steering allows for the tracking of 400-500 objects simultaneously while scanning at the same time, as well as filtering the image and detecting objects. The point cloud can also be output directly so that system developers can use the data for machine learning and edge-recognition algorithms in the sensor unit, and deliver that data to a central processing unit. Each radar channel is independent, so smaller arrays of transmitters and receivers can be used for different application requirements. For example, in driverless cars the performance requirements of a side-looking radar will be different from a forward-looking radar that is also looking for pedestrians and building a high-resolution map of the area around the vehicle. The specification for the low-end radar systems at Levels 3 and 4 autonomy is not yet agreed, say the radar makers, so flexibility is necessary. At the same time, higher levels of performance are needed for Level 5 systems, which have to handle any possible environment. Early versions of a 15 W radar imaging system will ship with 48 receivers and 32 transmitters for prototyping, while a 48/48 system is set for qualification to the ASIL B safety standards early next year. Networked radar Another approach has been to network the radar. To that end, a 36 W Ku-band (12-18 GHz) radar system has been designed to be less than 1.5 lb. It can be used on the ground to create an array that detects all the activity in the air. Linking this data back through the cloud, individual unmanned aircraft can be notified of objects around them, dramatically reducing the requirements for the aircraft. This works particularly well if corridors are defined for UAVs delivering packages, for example, and provides redundancy, avoiding the issue of a single point of failure in the aircraft. It also allows more complex, power- hungry algorithms to be used in cloud- based computing, while a GPU in the ground radar and a 100 Gbit/s Ethernet connection back to the cloud reduces the detection latency. A UAV can be notified with data over an LTE 4G cellular link with 35 bytes of data representing a detection, sent every eighth of a second. This requires only a bandwidth of a few kbytes/s over the LTE network. The radar uses two sets of independent phased array antennas built with standard parts, one for transmit and the other for receive to scan the sky. Software on the GPU tracks individual objects, filtering out any false positives, creating tracks and passing them eight times a second to the application software. This relatively lightweight radar can also be used on the UAV. This has to link to an inertial navigation system (INS) in the radar to provide the accurate relative position. If the pitch, roll and yaw are changing and the radar detects reflections, the INS is used to determine that reflections 1, 2 and 33 for example are actually the same object. Using the external INS of the aircraft can work, but that has a much longer latency. Adding time stamps to the external INS data can work but it isn’t fast enough for real-time detect-and- avoid applications. The advantage of this dual-use design on the ground and in the air is to give total airspace awareness, but it requires more infrastructure. It also addresses the issue of what needs to be detected, whether it’s a light aircraft or a small UAV, as the cloud-based algorithms are picking up every object in the sky. The mesh network can also be optimised for the environment on the ground. For example, the range of some units can be reduced to look down the urban canyons formed by large buildings, or the power can be increased with different antennas for a range of tens of kilometres in more open areas. Digital altimeters The lack of a specification for airborne radar systems in UAVs has led to another use for a collision detection radar, as a digital altimeter. A single- channel implementation in a single CMOS chip and a low-cost, low-power FPGA, can provide accurate height October/November 2018 | Unmanned Systems Technology Increasing the level of integration in standard chip-making processes is one way to reduce the cost of radar systems that can be used in UAVs (Courtesy of Ainstein)