Control Systems

Rob Slade

Designing a control system for a supersonic land speed record car poses some singularly challenging problems, not least: how do you test its response to inherently dangerous situations, such as the failure of a braking chute to deploy ? Add to that the necessity to keep track of the specifications of all the sub-systems, like aerodynamic control surfaces, cockpit instruments, fuel and throttle controls, etc, which can all be expected to change many times during the design process, yet which must all work together seamlessly when the car is running in anger. Whilst the "control system" does not actually control the car - that's the driver's job, as clearly specified in the FIA rules - it does have to monitor dozens of inputs and react accordingly to control parts of the car in response to the driver's inputs, or even take over control to bring the car to a safe halt in the event of an unsafe situation developing which is beyond the driver's influence. The control system is the collective name for the myriad of sensors which provide the required feedback on airspeed, wheel speed, cross-wind, angle of aerodynamic control surfaces (diffusers), steering input, etc., and the actuators which adjust the physical controls (eg: the jet engine's throttle body). In addition, the control system must have one or more "brains" (central processor) and electronic means to communicate with these sensors and actuators, which are distributed across the car. 

For Jetblack we are looking into using the concept of "model-based design" to specify and test the control system. A computer model of the car's dynamic behaviour can be created using sophisticated modern software packages; similar to the "physics engine" which is at the heart of computer game simulations of racing cars and aeroplanes, the physics model is a realistic representation of the response of the car to the effect of forces acting upon it, like the thrust of the jet and rocket engines, the drag of the air rushing past it, the side-forces due to crosswinds and steering, and the shocks fed through the suspension from the desert surface speeding beneath it. This model, which can be linked to a cockpit simulator and interacted with directly by a human driver, will allow us to run "what if" scenarios which we dare not test in reality, like the chute failure described above. If the control system or aspects of the car's design change during the engineering development phase (and they will, many times), the model can be updated and re-tested, validating that the control system is safe under all conceivable failure scenarios. As the design concept moves into the hardware phase it should be possible to create the software which runs in the on-board processor(s) directly from the validated physics model, and test the hardware "in the loop": the control system processor thinks it is interacting with a real car, but is actually receiving feedback from the model.