Why should car drivers love Kalman filtering?
23-03-2016 | By Paul Whytock
Fundamentally the Kalman filter is a sensor and data fusion algorithm. It uses the control inputs of a particular system combined with a substantial amount of measurements from sensors to rapidly create an estimate of the system's current operational state.
The reason why drivers should love it is simple enough, it increases the efficiency of advanced driver assistance systems (ADAS) and makes vehicle control operations like blind spot detection, stability and traction control, lane departure detection and automatic braking in emergency situations a lot safer and more effective.
Kalman filtering is particularly suitable for deployment in sensor-based ADAS as part of the radar tracker. Its task in these systems is to smooth the position and velocity data provided by the radar sensors and front-end digital signal processor (DSP) unit.
Specialist in low-power ASIC design and FPGA-based embedded systems, EnSilica, has just launched a Kalman Filter acceleration IP core for use in position awareness radar sensors for ADAS.
The company says that this IP core, which is part of EnSilica's automotive IP portfolio, provides speed improvement over software-only solutions of up to 10x.
The development of the company’s Kalman Filter acceleration IP core adheres to the guidelines required for integration with devices conforming to the ISO 26262 functional safety standard for road vehicles. It is also designed to handle both classical Kalman filtering (KF) and extended Kalman filtering (EKF). EKF is relevant when there is a non-linear relationship between the target’s Kalman state and the radar measurements.
The company says the compact low gate-count architecture of its Kalman Filter acceleration IP core enables the computationally intensive matrix operations involved in Kalman filtering to be offloaded from the CPU. It operates on distance, velocity and positional measurements and applies Kalman filtering so it can predict the target’s position in the next time interval. It combines the radar measurements with a dynamic motion model for enhancing the target position and velocity estimates with forward prediction, allowing false alarm measurements to be discarded.
Rapid refresh calculations
The system also provides a generic algorithm for combining measurements from different sensors into a single target track.
For a typical automotive radar system this Kalman Filter acceleration IP core can provide a system status refresh calculation in approximately 10µs. This enables a sizeable amount of target tracks to be maintained as this level of processing capability is short compared to a usual radar measurement cycle. The core undertakes five principle computational steps - setting the initial values, prediction of the state and error co-variance, computation of the Kalman gain, computation of the estimate and computation of the error co-variance - using floating point arithmetic in order to maintain numerical stability and provide equivalent results to a software based operation. RAM blocks of typically 8Kbits, depending on the matrix dimensions, are used to hold the computational matrices.