Unmanned Systems Technology 011 | C-Astral Bramor ppX | IMUs | Autonomous farming | UAV Turbines UTP50R | London Show report | Advanced materials | Las Vegas Expo report

38 Focus | Inertial measurement units Other sensors There are some cases where magnetometer information can be used for heading, orientation or spin rate but that can be very challenging in real-world conditions when dealing with the metal on the platform, such as in the motors. MEMS-based magnetometers with a U-shaped cantilever beam also provide a comparison with the gyros by checking how often the craft turns. MEMS-based pressure sensors can also be used as altimeters, again checking that the results of the IMU’s output are correct. Building an IMU Selecting the best type of sensor for a given application is a key part of building an IMU, and that can mean using several different types of sensor. In a high-end IMU for example, two different MEMS gyros are used on each of three axes, with each using different technologies to give different resolutions and dynamic ranges. The IMU then blends the data from each of those December/January 2017 | Unmanned Systems Technology sensors to give the result for that axis. This keeps the dynamic range very low to get better resolution. If the IMU is rotated too quickly, data is lost from the precision gyro, but the less accurate one has a higher dynamic range so the data is still captured. That is particularly useful for vehicles that need guidance but which are not turning or accelerating quickly. These high-end systems are providing a bias stability of 0.1°/hour and an angular random walk of 0.01°/route-hour. This level of performance is necessary in autonomous cars that need positional data that is accurate to a centimetre over a 500 m stretch, especially in an urban canyon where no navigation satellites are ‘visible’. The car needs to keep moving even if there is uncertainty over the position, as it cannot just stop in the middle of the road. Calibration Calibrating an IMU is an essential part of its development. To evaluate its response to acceleration and turns, the unit is tested on an arm that accelerates and rotates in two axes in a custom chamber. The resulting data, with temperature response and bias stability, is used to build a model of the unit’s performance, rather than just using a look-up table from the raw data. The model, built with second- or third-order polynomial equations or Kalman filters, is then used to provide the data for temperature compensation in the device. That means the linearity of the sensors is an essential parameter so that the model can be as accurate as possible. This feeds back into the design of the sensor’s structure to enhance the linearity of the response over other factors such as dynamic range. Temperature compensation is handled by signal processing within the IMU. It can be implemented by a dedicated digital signal processor (DSP), but the increasing performance of low-power embedded microcontrollers from ARM is providing sufficient performance. The latest Cortex-M4 processors with floating- point and DSP extensions are suitable for this, and are available from a range of chip suppliers. The microcontrollers can run a Coupling accelerometers, gyroscopes and GPS/GNSS data through an extended Kalman filter in an IMU provides position, velocity and attitude that can be orders of magnitude better than a standard navigation system (Courtesy of Vector Navigation) Selecting the best type of sensor is a key part of building an IMU, and that can mean using several different types of sensor