Post-doc project PILOT embeds executable physics models of wheel–soil interaction directly into the control system of heavy construction machines, enabling them to read the ground in real time and adapt their behaviour as terrain conditions change.
PILOT develops machine controllers in which a running physics model of wheel–soil interaction sits at the heart of the control loop. Sensor signals like wheel speed, drive torque, engine data, feed continuously into this embedded model, which infers ground properties such as softness and cohesion and updates the controller’s strategy in real time. When conditions change, the model detects it and adapts. Machine learning supports the estimation of parameters that cannot be measured directly, but the physical equations govern prediction and control.
A wheel loader working at a mine or construction site may cross firm gravel, wet clay, and loose sand within a single shift. Each terrain type demands a different driving strategy, but today’s machines use fixed controllers that cannot adapt. When the ground softens unexpectedly, wheel slip increases, fuel consumption rises, and the machine risks getting stuck.
PILOT’s approach is to place an executable physics model at the core of the control system. Grounded in soil mechanics and multibody dynamics, this numerical model describes how forces develop between a spinning wheel and deformable ground. It runs continuously alongside the machine, receiving sensor data and maintaining an up-to-date picture of the ground — its resistance to compression, its cohesion, how quickly traction is lost as slip increases. When conditions change, the model updates its parameters and passes a revised prediction to the controller, which adjusts wheel torque and speed to keep the machine near its optimal operating point. The goal is a wheel loader, and eventually other heavy off-road vehicles, that stays efficient and mobile whatever the terrain.
For further information, visit our homepage!
