Full-stack autonomous driving system spans diverse technological domains-including perception, planning, and control-that each require in-depth research. Moreover, validating such technologies of the system necessitates extensive supporting infrastructure, from simulators and sensors to high-definition maps. These complexities with barrier to entry pose substantial limitations for individual developers and research groups. Recently, open-source autonomous driving software platforms have emerged to address this challenge by providing autonomous driving technologies and practical supporting infrastructure for implementing and evaluating autonomous driving functionalities. Among the prominent open-source platforms, Autoware and Apollo are frequently adopted in both academia and industry. While previous studies have assessed each platform independently, few have offered a quantitative and detailed head-to-head comparison of their capabilities. In this paper, we systematically examine the core modules of Autoware and Apollo and evaluate their middleware performance to highlight key differences. These insights serve as a practical reference for researchers and engineers, guiding them in selecting the most suitable platform for their specific development environments and advancing the field of full-stack autonomous driving system.

Abstract Accurate vehicle speed estimation is essential for forensic traffic accident reconstruction, yet conventional video-based methods typically require prior camera calibration or spatial road data. This study introduces a modular framework that estimates vehicle speed from fixed-camera footage even when parts of the vehicle trajectory are completely occluded, without any prior information about the camera or accident scene. The framework integrates SiamRPN-based single-vehicle tracking with Kalman filtering to handle occlusions and uses Mask R-CNN instance segmentation to extract wheel centers. It then applies a smoothed unscented Kalman filter to predict occluded wheel coordinates and computes vehicle speed from the geometric cross-ratio—a perspective-invariant distance ratio—combined with known wheelbase specifications. Validation across three settings demonstrated high accuracy: simulations on straight and curved roads yielded maximum errors of 0.30 km/h and 1.70 km/h, respectively; real vehicle tests achieved errors of 0.41 km/h under constant-speed driving and 0.57 km/h during acceleration; and actual accident case studies produced errors of 0.49–3.05 km/h compared with field measurements. Unlike existing approaches that provide only point or interval estimates, the proposed framework generates continuous speed profiles throughout the accident sequence, enabling dynamic analysis of driver behavior such as acceleration, deceleration, and braking response. When vehicle specifications are available, this approach delivers legally defensible speed estimates from readily available video evidence, significantly improving the precision and practical utility of forensic accident reconstruction.

Accurate 3D multi-object tracking (MOT) is vital for autonomous vehicles, yet LiDAR and camera-based methods degrade in adverse weather. Meanwhile, Radar-based solutions remain robust but often suffer from limited vertical resolution and simplistic motion models. Existing Kalman filter-based approaches also rely on fixed noise covariance, hampering adaptability when objects make sudden maneuvers. We propose Bay es-4DR Track, a 4D Radar-based MOT framework that adopts a transformer-based motion prediction network to capture nonlinear motion dynamics and employs Bayesian approximation in both detection and prediction steps. Moreover, our two-stage data association leverages Doppler measurements to better distinguish closely spaced targets. Evaluated on the K-Radar dataset (including adverse weather scenarios), Bayes-4DRTrack demonstrates a 5.7% gain in Average Multi-Object Tracking Accuracy (AMOTA) over methods with traditional motion models and fixed noise covariance. These results show-case enhanced robustness and accuracy in demanding, real-world conditions.

Recently, 4D Radar has emerged as a crucial sensor for 3D object detection in autonomous vehicles, offering both stable perception in adverse weather and high-density point clouds for object shape recognition. However, processing such high-density data demands substantial computational resources and energy consumption. We propose SpikingRTNH, the first spiking neural network (SNN) for 3D object detection using 4D Radar data. By replacing conventional ReLU activation functions with leaky integrate-and-fire (LIF) spiking neurons, SpikingRTNH achieves significant energy efficiency gains. Furthermore, inspired by human cognitive processes, we introduce biological top-down inference (BTI), which processes point clouds sequentially from higher to lower densities. This approach effectively utilizes points with lower noise and higher importance for detection. Experiments on K-Radar dataset demonstrate that SpikingRTNH with BTI significantly reduces energy consumption by 78% while achieving comparable detection performance to its ANN counterpart (51.1% AP3D, 57.0% APBEV). These results establish the viability of SNNs for energy-efficient 4D Radar-based object detection in autonomous driving systems. All codes are available at https://github.com/kaist-avelab/k-radar.

Sensor fusion of camera, LiDAR, and 4-dimensional (4D) Radar has brought a significant performance improvement in autonomous driving. However, there still exist fundamental challenges: deeply coupled fusion methods assume continuous sensor availability, making them vulnerable to sensor degradation and failure, whereas sensor-wise cross-attention fusion methods struggle with computational cost and unified feature representation. This paper presents availability-aware sensor fusion (ASF), a novel method that employs unified canonical projection (UCP) to enable consistency in all sensor features for fusion and cross-attention across sensors along patches (CASAP) to enhance robustness of sensor fusion against sensor degradation and failure. As a result, the proposed ASF shows a superior object detection performance to the existing state-of-the-art fusion methods under various weather and sensor degradation (or failure) conditions. Extensive experiments on the K-Radar dataset demonstrate that ASF achieves improvements of 9.7\% in $AP_{BEV}$ (87.2\%) and 20.1\% in $AP_{3D}$ (73.6\%) in object detection at IoU=0.5, while requiring a low computational cost. All codes are available at https://github.com/kaist-avelab/k-radar.

Lane detection algorithm (LDA) is a crucial and necessary component for autonomous vehicles to ensure safe driving in various environments. Deep learning-based lane detection algorithms (DL-LDAs) have gained significant attention recently, and there have been a number of DL-LDAs, introduced in the literature, showing a continuous performance improvement in lane detection. In general, DL-LDAs are composed of pre-processing, lane feature extraction, lane detection head, and an optional lane fitting. For a systematic overview of various DL-LDAs, we provide detailed explanations for each functional component of DL-LDAs. Moreover, this paper presents the first survey to comprehensively analyze various DL-LDAs using camera and LiDAR, such as 2D (2-Dimensional) and 3D DL-LDAs using camera images and DL-LDAs using LiDAR point cloud or sensor fusion. In addition to the analysis, we present recent public lane detection benchmarks for DL-LDAs and discussions concerning technical issues that need to be addressed in future DL-LDA studies.

Focusing on the strength of 4D (4-Dimensional) radar, research about robust 3D object detection networks in adverse weather conditions has gained attention. To train such networks, datasets that contain large amounts of 4D radar data and ground truth labels are essential. However, the existing 4D radar datasets (e.g., K-Radar) lack sufficient sensor data and labels, which hinders the advancement in this research domain. Furthermore, enlarging the 4D radar datasets requires a time-consuming and expensive manual labeling process. To address these issues, we propose the auto-labeling method of 4D radar tensor (4DRT) in the K-Radar dataset. The proposed method initially trains a LiDAR-based object detection network (LODN) using calibrated LiDAR point cloud (LPC). The trained LODN then automatically generates ground truth labels (i.e., auto-labels, ALs) of the K-Radar train dataset without human intervention. The generated ALs are used to train the 4D radar-based object detection network (4DRODN), Radar Tensor Network with Height (RTNH). The experimental results demonstrate that RTNH trained with ALs has achieved a similar detection performance to the original RTNH which is trained with manually annotated ground truth labels, thereby verifying the effectiveness of the proposed auto-labeling method. All relevant codes will be soon available at the following GitHub project: https://github.com/kaist-avelab/K-Radar

In a multipath channel environment, classifying non-line-of-sight (NLOS) Global Navigation Satellite (GNSS) satellites and compensating multipath ranging error (MRE) is the most important task for improving GNSS positioning accuracy in urban areas. Recently, Signal-to-noise ratio (SNR), pseudorange, and other measurements have been used to classify NLOS satellites, but these measurements have limited representation of NLOS channel characteristics. In this paper, we propose a Multipath error Compensation Network (MPCNet) that uses an Autocorrelation function (ACF) output and 3D Geographic Information System (GIS) as inputs to classify NLOS satellites and compensate for MRE. MPCNet is composed of two heads for each task and a shared network that learns relevant information about the multipath channel environment from the ACF output. The performance evaluation of MPCNet was performed in a real urban environment, and the NLOS classification accuracy was compared with that of conventional deep learning-based NLOS classifiers, and the positioning performance of conventional positioning methods was also compared. MPCNet showed an NLOS classification performance of about 97% and an improvement in positioning accuracy of about 57% compared to conventional positioning methods, demonstrating that it is a robust and accurate MRE compensation technique in multipath environment.

Recent works have shown the superior robustness of four-dimensional (4D) Radar-based three-dimensional (3D) object detection in adverse weather conditions. However, processing 4D Radar data remains a challenge due to the large data size, which require substantial amount of memory for computing and storage. In previous work, an online density reduction is performed on the 4D Radar Tensor (4DRT) to reduce the data size, in which the density reduction level is chosen arbitrarily. However, the impact of density reduction on the detection performance and memory consumption remains largely unknown. In this paper, we aim to address this issue by conducting extensive hyperparamter tuning on the density reduction level. Experimental results show that increasing the density level from 0.01% to 50% of the original 4DRT density level proportionally improves the detection performance, at a cost of memory consumption. However, when the density level is increased beyond 5%, only the memory consumption increases, while the detection performance oscillates below the peak point. In addition to the optimized density hyperparameter, we also introduce 4D Sparse Radar Tensor (4DSRT), a new representation for 4D Radar data with offline density reduction, leading to a significantly reduced raw data size. An optimized development kit for training the neural networks is also provided, which along with the utilization of 4DSRT, improves training speed by a factor of 17.1 compared to the state-of-the-art 4DRT-based neural networks. All codes are available at: https://github.com/kaist-avelab/K-Radar.

Among the challenges in the recent research of end-to-end (E2E) driving, interpretability and distribution shift in the simulation-to-real (Sim2Real) have drawn considerable attention. Because of low interpretability, we cannot clearly explain the causal relationship between the input image and the control actions by the network. Moreover, the distribution shift problem in Sim2Real degrades the driving performance of the policy in the realworld deployment. In this paper, we propose a segmentation-based classwise disentangled latent encoding algorithm to cope with the two challenges. In the proposed algorithm, multi-class segmentation transfers RGB images in both simulation and real environments to the same domain, while preserving the necessary information of objects of primary classes, such as pedestrian, road, and cars, for driving decisions. Besides, in the class-wise disentangled latent encoding, segmented images are encoded to a latent vector, which improves the interpretability significantly, since the state input has a structured format. The interpretability improvement is testified by the t-stochastic neighbor embedding, image reconstruction and the causal relationship between the real images and the control actions. We deploy the driving policy trained in the simulation directly to an autonomous vehicle platform and show, to the best of our knowledge, the first demonstration of the RL-based E2E autonomous in various real environments.

Unlike RGB cameras that use visible light bands (384∼769 THz) and Lidars that use infrared bands (361∼331 THz), Radars use relatively longer wavelength radio bands (77∼81 GHz), resulting in robust measurements in adverse weathers. Unfortunately, existing Radar datasets only contain a relatively small number of samples compared to the existing camera and Lidar datasets. This may hinder the development of sophisticated data-driven deep learning techniques for Radar-based perception. Moreover, most of the existing Radar datasets only provide 3D Radar tensor (3DRT) data that contain power measurements along the Doppler, range, and azimuth dimensions. As there is no elevation information, it is challenging to estimate the 3D bounding box of an object from 3DRT. In this work, we introduce KAIST-Radar (K-Radar), a novel large-scale object detection dataset and benchmark that contains 35K frames of 4D Radar tensor (4DRT) data with power measurements along the Doppler, range, azimuth, and elevation dimensions, together with carefully annotated 3D bounding box labels of objects on the roads. K-Radar includes challenging driving conditions such as adverse weathers (fog, rain, and snow) on various road structures (urban, suburban roads, alleyways, and highways). In addition to the 4DRT, we provide auxiliary measurements from carefully calibrated high-resolution Lidars, surround stereo cameras, and RTK-GPS. We also provide 4DRT-based object detection baseline neural networks (baseline NNs) and show that the height information is crucial for 3D object detection. And by comparing the baseline NN with a similarly-structured Lidar-based neural network, we demonstrate that 4D Radar is a more robust sensor for adverse weather conditions. All codes are available at https://github.com/kaist-avelab/k-radar.

BPSK modulated GPS L1 CA signal is the most widely used GNSS signal to date, and the first path detection (FPD) of the conventional GPS L1 CA signals is the most challenging problem to ensure reliable GPS positioning in multipath environments. In this paper, we propose an FPD network (FPDN) based on multi-layer perceptron (MLP)-Mixer to extract the first path from the discrete autocorrelation function (ACF) output accurately with low computational cost. In addition, the proposed FPDN is useful in practice because it is robust to noise and achieves a high FPD performance without any prior assumption on the number of total incoming multipath, which is required for conventional signal processing-based FPD techniques. We compare the performance of the proposed FPDN to that of diverse conventional techniques, such as techniques based on narrow correlator, super-resolution, and some widely used CNNs such as VGGNet, ResNet, and U-Net, through simulations and field tests. As demonstrated, the proposed FPDN outperforms all of the compared FPD techniques in terms of the computational cost and accuracy for wide range of carrier-to-noise (C/N0) ratios.

There has been an increasing level of demand for faster, safer and greener transportation systems with higher levels of capacity and convenience, though the implementation of transportation systems overall is often restricted by geographical limitations, presenting a challenge to scientists and engineers in the field. However, we have been witnessing the evolution of the transportation systems over the last few decades, and at present we are facing a new era of intelligent transportation systems (ITS) empowered by artificial intelligence (AI) technologies. There have been classification, deep learning, and reinforcement learning techniques, to name a few, which collectively have enabled almost all technical elements of the ITS. For example, autonomous vehicle technologies are now mature enough to introduce self-driving cars, taxis, buses, and trucks on the roads and streets; traffic signals are controlled by AI-based systems for far more enhanced traffic efficiency; and machine learning based on big data is improving the operational performance of transportation systems to the next level of safety, efficiency, and sustainability.

Most Conventional traffic light control (TLC) techniques do not provide enough efficiency to control dynamic traffic situations in real-time. Recently, DQN (Deep Q Network) algorithm is considered for TLC at the intersection because of its optimization technique for complex problems, where key features of the intersection traffic, such as vehicle positions and velocities, are obtained from the intersection by the camera installed at well above the ground. However, the general DQN-based TLC algorithms have failed to utilize the fact that vehicle trajectories are continuous, which can be very useful in sensing real-time traffic. To utilize the continuous vehicle motion for TLC improvement, we propose DRQN-TLC (Deep Recurrent Q Network for TLC) algorithm that is based on LSTM (Long-Short Term Memory) with DQN. The superior performance of the proposed algorithm is demonstrated with the simulation; the proposed algorithm reduces the average traveling time by 23% and the overall vehicle waiting time by 10% when compared with the general DQN-based TLC algorithm.

In the cooperative vehicular positioning networks (CVPN), non-line-of-sight (NLOS) delay in the vehicle-to-vehicle (V2V) ranging and malicious attacks, such as location spoofing and the manipulation of V2V ranging, can be significant threats to vehicle safety. However, the resulting observation from a vehicle in the event of any of the threats is the same; difference observation between the measured range and the Euclidian distance based on vehicles' shared location coordinates. In this paper, we propose a decentralized malicious vehicle detection technique for cooperative positioning technique (CPT), in which each vehicle evaluates the reliability of neighboring vehicles based on the observed difference and share the reliability with each other, so that each vehicle in the CVPN can collect the reliability of neighboring vehicles evaluated by their own neighboring vehicles. However, it is found that a malicious vehicle can make another attack; distribution of distorted reliability of its neighboring vehicles. To cope with the malicious attacks and NLOS delay effectively, we propose the weighted reliability and demonstrate that the proposed technique can detect malicious attacks (location spoofing, ranging manipulation, message distortion) and distinguish malicious vehicles from vehicles with a NLOS link. With Monte Carlo simulations using V2V radio parameters determined empirically, we demonstrate that the proposed technique achieves higher enhancement and superior robustness to malicious attacks than the conventional CPTs.

Due to secondary code and AltBOC modulation,the primary code acquisition of the Galileo E5 signal can be complicated and requires additional hardware sources and algorithmic complexity in a receiver. In this paper, we propose a fast primary code acquisition technique for the Galileo E5 signal to reduce both hardware and algorithm complexities, while achieving a similar or better performance in receiver operating characteristic (ROC) and mean acquisition time (MAT), respectively. The proposed technique employs a sub-Nyquist sampling scheme followed by a sample compression scheme to cause a complete aliasing between the spectra of E5a and E5b signals to minimize the signal bandwidth and to reduce the number of code phase hypothesis to search. We demonstrate with numerous Monte Carlo simulations that the MAT of the proposed technique is about a half of the conventional AltBOC acquisition techniques.

The accuracy of the observed time difference of arrival (OTDOA) in the long-term evolution (LTE) systems depends on the accuracy of the time of arrival (TOA) measurements, which are often corrupted by various errors caused by non-light-of-sight propagation, multipath interference, noise, and path detection techniques. Furthermore, signal bandwidth, channel condition, distance from the evolved node-B, and scatterer distribution are the affecting parameters on the OTDOA accuracy. Since the user equipment obtains the most accurate TOA from the resolved first arrival path (R-FAP), understanding errors of the TOA and OTDOA from the R-FAP is necessary to develop OTDOA positioning techniques. In this paper, we develop theoretical expressions for the TOA and OTDOA error distributions of the R-FAP for outdoor multipath environments by integrating theoretical models of the errors expressed with the affecting parameters, and theoretical expressions are verified with numerous Monte Carlo simulations. In addition, we propose an LTE OTDOA positioning technique that compensates the mean TOA offset in the TDOA measurements before applying a positioning algorithm, and we demonstrate the performance improvement using Monte Carlo simulations. In this paper, we do not include OTDOA errors due to the network synchronization and intercell interference.

A magnetometer is an emerging indoor positioning technique that has a strong advantage in infrastructure requirement. However, due to the cheap magnetometers used in mobile devices, magnetometer based indoor positioning technique suffers from the bias of the magnetometer. In this paper, we propose magnetometer measurement difference (MMD)-based technique for Bayesian filter to mitigate this problem. The proposed technique is much more robust to the bias of magnetometer than the conventional technique when the triaxis of magnetometer in a user device is tilted away from the triaxis of magnetic field map. We provide theoretical analysis and simulation results to compare the performance of the proposed MMD-based technique to the conventional magnetometer-based positioning technique.

Successful and fast Global Navigation Satellite System (GNSS) positioning in indoor environments can enable many location based services (LBS). However, fast indoor GNSS positioning has been one of the biggest challenges for GNSS receivers due to the huge computational cost. To detect weak GNSS signals in indoor environments, a GNSS receiver should perform numerous correlations with a longer coherent integration interval for a denser Doppler frequency search, which is computationally too expensive. For a fast and low computational weak GNSS signal detection, we propose the synthesized Doppler frequency hypothesis testing (SDHT) technique that, utilizing the test results of only sparse Doppler frequency hypotheses, can estimate the test results of entire Doppler frequency hypotheses with small computations. We provide theoretical performance analysis of the proposed technique and demonstrate that the proposed technique reduces the computational cost for weak GNSS signal acquisition significantly and achieves faster signal acquisition than conventional techniques.

The position and velocity information of high-speed trains (HSTs) are essential to passenger safety, operational efficiency, and maintenance, for which an accurate navigation system is required. In this paper, we propose a two-stage federated Kalman filter (TS-FKF) for an HST navigation system that uses multi-sensors, such as tachometer, inertial navigation system, differential GPS, and RFID, with a feedback scheme. However, the FKF with a feedback scheme often shows severe performance degradation in the presence of undetected large sensor errors. Tachometers often have large slip or slide errors during the train's acceleration, deceleration, and moving along a curved railway, and there are significant performance differences between different sensors. To make the proposed system robust to these errors, we propose a slip and slide detection algorithm for the tachometer and an adaptive information-sharing algorithm to deal with a large tachometer error and performance difference between sensors. We provide theoretical analysis and simulation results to demonstrate the performance of the proposed navigation system with the proposed algorithms.

Due to the longer PRN code employed in next-generation GNSS (Global Navigation Satellite System), receivers in the signal acquisition process should test a larger number of code phase hypotheses. Recently, to reduce acquisition time and computational complexity, TDCC (Two-dimensional Compressed Correlator) is introduced, and this technique is different from the conventional double dwell search technique in terms of parameters used to reduce acquisition time and computational complexity. Studies for optimizing the performance of TDCC have not been introduced yet, therefore, in this paper, detection thresholds of TDCC are investigated to optimize its acquisition performance. The optimal detection thresholds minimizing the MAT (Mean Acquisition Time) and MAC (Mean Acquisition Computation) are numerically evaluated and analyzed for widely used detection strategies in its serial and parallel search schemes. It is demonstrated that the minimized MAT and MAC are much smaller than the MAT and MAC using a detection threshold based on CFAR (Constant False Alarm Rate).

Due to the longer spreading code used for the next generation GNSS (Global Navigation Satellite System) signals, receivers have to spend longer time or require larger amount of hardware resources for signal acquisition. Since many recent GNSS receivers use DSP (Digital Signal Processor) to realize parallel signal acquisition scheme in the frequency domain, this paper proposes FFT-based TDCC (Two-Dimensional Compressed Correlator) with which coherently compressed hypotheses are tested using a reduced number of IFFT points in the 1st stage, and the individual hypotheses that construct the compressed hypothesis found in the 1st stage are tested using the conventional parallel search scheme in the 2nd stage. The performance of the proposed technique is demonstrated with numerous Monte Carlo simulations and a comparison to the conventional FFT-based search technique is provided. The results show that the proposed technique requires lower computation and has lower mean acquisition time than the conventional FFT-based search technique for moderate and high C/N0(Carrier-to-Noise Density Ratio) GNSS signals.

Successful position fix in harsh environments such as indoors and dense urban canyons is a strongly required capability for an assisted global positioning system (A-GPS) receiver. In recently developed cellular networks, receiving fine time assistance and maintaining high-frequency accuracy using downlink measurements are not possible for A-GPS receivers, since node-Bs are asynchronous and are not equipped with a source for precise time and frequency. In this paper, we propose a correlator-based fast multi-satellite maximum likelihood (MSML) algorithm, for A-GPS receivers in asynchronous networks, that achieves fast acquisition utilizing fast computation techniques. From numerous Monte Carlo simulations, it is demonstrated that the proposed fast MSML algorithm, when compared with conventional correlator-based acquisition techniques used in standalone GPS and A-GPS receivers, provides higher detection sensitivity for weak signals in the presence of other strong signals by removing strong inter-satellite interference (ISI).

Acquisition of an incoming long pseudo-noise (PN) code signal requires a fast hypothesis testing function for a large number of code phase hypotheses. In addition, when a transmitter is moving at a high speed, the hypothesis testing function needs to search for the signal in a 2-Dimensional (2D) search space that includes all possible combinations of code phase hypothesis and Doppler frequency hypothesis. Since a receiver has limited hardware resources (in terms of number of correlators and computational capacity) in practice, fast PN code acquisition is not an easy goal to achieve. In this paper, we propose a double dwell search scheme, where a 2D compressed correlator (TDCC) tests a number of coherently combined neighboring code phase hypotheses and Doppler frequency hypotheses at a time in the 1st dwell search, and all individual neighboring hypotheses found in the 1st dwell search are tested in the 2nd dwell search using a conventional correlator. We present theoretical performance analysis of the proposed technique and Monte Carlo simulation results to demonstrate the performance of the proposed technique and to compare it to the conventional double dwell search technique.

Increasing the dwell time in two-dimensional frequency-time hypothesis testing is, in practical terms, one of the most effective ways for Assisted Global Navigation Satellite Systems (A-GNSS) and GNSS receivers to achieve higher sensitivity. In an asynchronous cellular network, however, a mobile terminal may have a non-negligible unknown clock drift rate error originating from the received cellular downlink signal. In such a case, increasing the dwell time may not necessarily result in the expected sensitivity improvement. In addition, a mobile terminal in a rich multipath environment may experience jitters in the code phase of the resolved first arrival path due to short-delay multipaths, which also degrades the sensitivity. In this paper, new decision variables using a lone or a pair of adjacent H_1 cells for code phase hypothesis testing and clock drift rate hypothesis testing are proposed to cope with the unknown code phase drift rate error and the effect of code phase jitter in a parallel acquisition system. The statistics of the proposed decision variables are analyzed in a Rayleigh fading channel, and the performances of the proposed decision variables are compared with that of the conventional decision variable.

Gaussian scatterer distribution model (GSDM) is one of the most interesting geometrically based scatterer distribution models for spatial and temporal properties of wireless channels in most multipath environments. The GSDM assumes a circular scattering region around a mobile station (MS), and the scatter density decreases with the distance from the MS. In this paper, the time of arrival (TOA), angle of departure (AOD), and joint TOA/AOD probability density functions (pdfs) of down link are derived for the GSDM. Based on these pdfs, the TOA and AOD pdfs of first arrival path are analyzed as they are particularly important to radiolocation technologies. A closed-form expression for the TOA pdf and an approximate expression for the AOD pdf of the first arrival path are obtained. To validate the closed-form expressions, comparisons to simulated normalized histograms obtained from Monte Carlo trials are included. The pdfs derived in this paper provide substantial insight into the statistical properties of wireless channels and first arrival path in multipath environments.