Unmanned Systems Technology 020 | Alpha 800 I Additive Manufacturing focus I USVs insight I Pegasus GE70 I GuardBot I AUVSI Xponential 2018 show report I Solar Power focus I CUAV Expo Europe 2018 show report

66 Digest | GuardBot are off the central axis of the sphere and driven by their own motor. The main pendulum is driven along the axis of the ball’s rotation, and the second is perpendicular to the first. That allows the mass of the second pendulum to swing to the left or right to provide the steering. The dynamic model shows that there are nine possible states of operation to consider. These are related to the rotational position, velocity and acceleration of the shell, the drive pendulum and the steering pendulum. There are only two inputs to the dynamic model – the drive motor’s torque and the steering motor’s torque. The internal coupling of the pendulums with the spherical shell, the movement of both pendulums and the constraints imposed by the non-slip ground contact mean the system’s dynamic model is quite complicated, but was simplified and computed using the Mathematica symbolic solver software tool. The modelling showed that the robot is controllable via the two motor torques only if the initial yaw position of the system is not considered, and there is no direct control of the direction in which the robot is pointing without travelling forwards and turning. That is the same as a typical car. That means the system can be controlled using direct measurements of the drive motor’s velocity, the steering motor’s velocity, the steering motor’s position, pendulum pitch and roll (relative to gravity) and the initial yaw. Once the system had been modelled it was simulated in software using the Matlab tool. A linear feedback system was implemented around the point where the drive pendulum hangs directly downwards and the steering pendulum in the neutral position. The motor dynamics that relate torques to motor voltages were also included in the models to simulate motor voltage control. A control system was built using a technique called a linear quadratic regulator (LQR), with an internal Kalman filter to improve the internal state estimates of the system. The LQR automatically determines feedback gains based on the desired level of accuracy for each state. The emphasis was on the control of the sphere’s velocity, with less emphasis on the exact acceleration profile used to achieve it. The Kalman filter is designed to handle the variations in the accuracy across all the states to provide the optimum result. In the simulation, the combination of the LQR and Kalman filter were used to generate the configuration data to specify the motors, the expected motor torques and even tune the controller to get the best performance. The system was then implemented in hardware using two dynamic DC brushless motors from Maxon Motors, which provide 6.2 Nm of torque to the pendulums to provide acceleration and deceleration, although the torque can be as high as 23 Nm to tackle slopes and obstructions, for example to get over sleepers on a railway. It is this ability to provide higher torque to the drive pendulum to move the centre of mass, combined with the traction from the cover of the sphere, that also allows the GuardBot to move in difficult terrain such as snow and mud. Each motor is also connected to a planetary gearhead and rotary encoder. These give detailed position and velocity information direction to the internal controller to provide closed-loop control of the velocity, as opposed to the open- loop control of the pendulums, with direct connection to the motors’ quadrature encoders. The velocity data is then fed back to the controller via an analogue voltage signal. The controller also has direct torque control that can automatically compensate for changes in velocity. This data is fed to a central processor that also collects data from an internal accelerometer that provides roll-pitch- yaw information, and then issues the commands to the two motors’ controllers via analogue pulse width modulation signalling. Communication with the control unit is through a digital serial link. Radio system The link back to an operator is via an STx transmitter from Integrated Microwave Technologies. It uses a coded orthogonal frequency domain modulation (COFDM) protocol that supports HD video streams compressed using the H.264 standard. The radio delivers its HD/SD output at 20 or 250 mW, depending on range requirements, and occupies less than 5 cu in in the GuardBot. It supports June/July 2018 | Unmanned Systems Technology The GuardBot can operate with different payloads such as HD cameras or chemical sensors