Abstract:Accurate targets detection and tracking with mmWave radar is a key sensing capability that will enable more intelligent systems, create smart, efficient, automated system. This paper proposes an end-to-end detection-estimation-track framework named MNOMP-SPA-KF consisting of the target detection and estimation module, the data association (DA) module and the target tracking module. In the target estimation and detection module, a low complexity, super-resolution and constant false alarm rate (CFAR) based two dimensional multisnapshot Newtonalized orthogonal matching pursuit (2D-MNOMP) is designed to extract the multitarget's radial distances and velocities, followed by the conventional (Bartlett) beamformer to extract the multitarget's azimuths. In the DA module, a sum product algorithm (SPA) is adopted to obtain the association probabilities of the existed targets and measurements by incorporating the radial velocity information. The Kalman filter (KF) is implemented to perform target tracking in the target tracking module by exploiting the asymptotic distribution of the estimators. To improve the detection probability of the weak targets, extrapolation is also coupled into the MNOMP-SPA-KF. Numerical and real data experiments demonstrate the effectiveness of the MNOMP-SPA-KF algorithm, compared to other benchmark algorithms.
What problem does this paper attempt to address?
This paper attempts to solve several key problems in millimeter - wave radar (mmWave radar) for multi - target detection and tracking:
1. **Weak target detection**: In a strong background clutter environment, the signals from weak targets (such as pedestrians with small radar cross - section RCS) in the millimeter - wave radar reflection signals may be masked by the reflection waves from strong targets (such as buildings and trucks), resulting in these weak targets not being effectively detected.
2. **False alarm management**: Traditional detection algorithms are prone to generate a large number of false alarms, which not only increase the computational burden of the system but also affect the overall performance. Especially when the weak target is close to the strong target, the traditional constant false - alarm rate (CFAR) detector may miss the weak target.
3. **Multi - target data association**: In a multi - target scenario, how to effectively match the sensor measurement results with the existing targets and establish new trajectories is a complex problem. Traditional nearest - neighbor (NN) algorithms and probabilistic data association (PDA) algorithms perform poorly in low signal - to - noise ratios or cluttered environments, especially when dealing with multiple targets.
To solve these problems, the paper proposes an end - to - end joint detection - estimation - tracking framework, called **MNOMP - SPA - KF**, which specifically includes the following modules:
- **Target detection and estimation module**: A two - dimensional multi - snapshot normalized orthogonal matching pursuit (2D - MNOMP) algorithm with low complexity, super - resolution and constant false - alarm rate (CFAR) characteristics is used to extract the radial distance, velocity and azimuth of multiple targets from the baseband signal.
- **Data association module**: The sum - product algorithm (Sum Product Algorithm, SPA) is used, combined with the radial velocity information, to obtain the association probability between the existing targets and the measurement results.
- **Target tracking module**: The Kalman filter (Kalman Filter, KF) is used to achieve target tracking, taking advantage of the asymptotic distribution characteristics of the estimators.
In addition, in order to improve the detection probability of weak targets, the framework also introduces the extrapolation method. Numerical experiments and real - data experiments verify the effectiveness of the MNOMP - SPA - KF algorithm, which shows better performance compared to other benchmark algorithms.
### Formula summary
1. **Baseband signal model**:
\[
Y_{n,m,l}(t)=\sum_{k = 1}^{K(t)}\gamma_k(t)e^{j(n - 1)\frac{4\pi}{c}\mu T_sr_k(t)}e^{j(m - 1)\frac{4\pi}{c}f_cT_rv_k(t)}e^{j(l - 1)\frac{2\pi}{c}f_cd\sin\theta_k(t)}+\epsilon_{n,m,l}(t)
\]
where:
- \(n\) is the fast - time domain index
- \(m\) is the slow - time domain index
- \(l\) is the spatial domain index
- \(N\), \(M\) and \(L\) are the measurement numbers in the fast - time domain, slow - time domain and spatial domain respectively
- \(K(t)\) is the number of targets
- \(\gamma_k(t)\) is the complex amplitude of the \(k\) - th target
- \(r_k(t)\), \(v_k(t)\) and \(\theta_k(t)\) are the radial distance, radial velocity and azimuth of the \(k\) - th target respectively
- \(c\) is the speed of electromagnetic waves
- \(\mu\) is the frequency - modulation slope
- \(T_s\) is the sampling interval
- \(f_c\) is the carrier frequency
- \(T_r\) is the frequency - modulation period
- \(d\) is the antenna element spacing
- \(\epsilon_{n,m,l}(t)\) is the additive noise, assumed to be independent and identically distributed (i.i.d.)
2. **State - transfer model**:
\[
x_k(t)=Ax_k(t - 1)+w_k
\]