A Distributed Sensing System for Vehicles State Estimation

This paper presents the design of a distributed sensing system that uses an Extended Kalman Filter (EKF) to fuse measurements, so that automotive vehicle states can be estimated for use by a Semi-active Suspension control system. To improve ride comfort and handling quality, relative displacements and velocities of suspension systems are estimated. To control the stability of vehicles, roll, yaw, and pitch must also be determined. The designed (EKF) uses easily accessible measurements such as accelerations and body’s angular velocities. These measurements are provided by 8 accelerometers and an Inertial Measurement Unit (IMU). The accelerometers are strategically mounted on the two ends of each individual shock absorber (damper). The IMU was mounted near the vehicle’s center of gravity. Computer simulations and experiments were conducted for full vehicle state estimation of a 1993 Toyota Tercel equipped with the above mentioned sensor suite. Results show that except relative displacements, all states of the automobile’s semi-active suspension systems can be estimated using this set of sensors. The designed EKF works well despite not knowing accurate information about road inputs, external disturbances and car characteristics such as moments of inertia, mass, and equivalent spring and damping coefficients. Both simulation results and experimental results show the effectiveness of the designed EKF in estimating the required states. INTRODUCTION During last two decades, active suspension systems and semi-active dampers have been developed to improve performance parameters corresponding to terrain vehicles. Various types of controllers including linear/non-linear optimal and robust strategies have been proposed by researchers [1]. While active and semi-active damper technologies offer new opportunities to improve ride comfort, road handling and vehicle stability, the primary challenge of providing required control system feedbacks remains. Many proposed control laws have assumed that all the system states including suspension systems deflection and its rate, velocity of wheel hubs and tire deflections are completely and accurately measurable, which is not always the case [15]. Velocity cannot be measured directly because velocity sensors must have a stationary reference space which is not applicable in automotive applications. Common techniques integrate accelerometer signals attached to the ends of each shock to achieve velocity information. Performance of low-cost automotive grade accelerometers is limited by high level noise (typically 50-1000 Hz g / μ ) and drift due to biases and gravitational effects. Passing the integrated signal through a high-pass filter is not always adequate for obtaining accurate velocity data. To make the various control methodologies realizable, a filter and/or observer is required to produce accurate state estimates. In [6] a modified Luenberger observer was presented to estimate semi-active suspension system states for a quarter