Kyounghwan An

High-Level Data Fusion Based Probabilistic Situation Assessment for Highly Automated Driving

A primary challenge of automated driving systems is the task of a situation assessment. This paper presents a high-level data fusion based probabilistic situation assessment method which is capable of assessing a current traffic situation and giving a recommendation about driving behaviors. The proposed method consists of two steps: high-level data fusion and probabilistic situation assessment. The high-level data fusion, designed to provide a better understating of observed situations, produces a local dynamic road map by integrating all dynamic entities with a high-precision static road map. The probabilistic situation assessment estimates threat levels of each lane as the probability of the lane state through the use of independent local experts based on the local dynamic road map. The recommendations for behavior decision are determined by filtering out noises resulting from object tracking even though a tracking module misses objects or detects wrong objects a lot, but immediately. The method is implemented in an open-source robot operating system to provide a reusable and hardware independent software platform, and verified and evaluated through in-vehicle tests on real highways in real-time operation.

Situation assessment and behavior decision for vehicle/driver cooperative driving in highway environments

This paper proposes an automated system with respect to situation assessment and behavior decision not only for cooperative driving between a driver and the system but also highly automated driving in highway environments. The proposed system includes three main parts: (1) high-level data fusion to produce a better understanding of the observed situation, (2) distributed reasoning based situation assessment to evaluate the current situation in the safety aspect and to recommend actions, and (3) behavior decision to determine collision-free and goal-directed maneuvers for vehicle/driver cooperative and highly automated driving in highway environments. The system is verified that it can work properly for cooperative and highly automated driving through in-vehicle tests in several scenarios similar to real highway environments.

Automatic Valet Parking System incorporating a nomadic device and parking servers

Automatic Valet Parking System (AVP) is a system that supports autonomous maneuvering from a traffic lane to free parking slot. In this paper, we propose a hybrid framework using both in-vehicle and infrastructure based system for detecting environment and controlling the vehicle. Using the system, drivers can get off their car, and can monitor/control them in a safe and comfort way with their nomadic device.

Design of query language for location-based services

Recently, the need for LBS (Location Based Services) is increasing due to the widespread of mobile computing devices and positioning technologies. In LBS, there are many applications that need to manage moving objects (e.g. taxies, persons). The trajectories or the positions of moving objects are displayed on the map by using GIS. However, it is hard to use GIS or traditional relational database systems to manage moving objects. Modeling consistent information about the location of continuously moving objects and processing motion-specific queries is challenging problem. The previous studies suggested several query languages to retrieve moving objects. However, they do not propose DDL (Data Definition Language) and do not support full functions that are necessary. In this paper, we formally define a data model and data type for moving objects and propose MOQL (Moving Objects Query Language) which is a convenient interface and tool for developers. MOQL has the following features. First, it supports DDL to insert/delete/update the positions of moving objects. Second, it can be used to retrieve the trajectories of moving objects. Third, it defines several functions to manage spatial or temporal properties of moving objects.

A Log-Based Cache Consistency Control of Spatial Databases in Mobile Computing Environments

In mobile client/server computing environments, mobile clients make access to their server to get interested data and then are disconnected because of high cost of wireless communication. Mobile clients usually keep their own local copies in order to reduce the overhead of communicating with the server. The updates of the server database sometimes are subject to leading to invalidation of the cached map in mobile clients. However it is not efficient to resend the entirely copied map from the server to mobile clients for solving invalidation. This paper proposes a log-based update propagation method to propagate the servers update into its corresponding mobile clients by sending only update logs. The log-based update propagation scheme raises new issues as follows. First, the continuously growing of update logs downgrades the speed of searching for the relevant log data for a specific client. Second, there is considerable overhead of transmitting the update logs into mobile clients by using wireless communication. To solve these problems, we define unnecessary logs and then suggest methods to remove the unnecessary logs.

Decision-Making Framework for Automated Driving in Highway Environments

This paper presents a decision-making framework for automated driving in highway environments. The framework is capable of reliably, robustly assessing a given highway situation (with respect to the possibility of collision) and of automatically determining an appropriate maneuver for the situation. It consists of two main components: situation assessment and strategy decision. The situation assessment component utilizes multiple complementary “threat measures” and Bayesian networks in its calculations of “threat levels” at the car and lane level to evaluate the possibility of collisions for a given highway traffic situation. The strategy decision component, designed to generate goal-directed and collision-free behaviors, automatically determines an appropriate maneuver in a given highway situation via a hierarchical state machine—such a machine both reduces the complexity of and extends a strategy model. The types of maneuver determined by the component include both simple maneuvers, such as slowing down to avoid collision with a vehicle in front, and complex maneuvers, such as lane changes and overtaking. The presented decision-making framework is tested and evaluated—both on a closed high-speed test track in simulated traffic with various driving scenarios and on public highways in real traffic through in-vehicle testing—to verify that it can provide sufficiently reliable performance for automated driving in highway environments in terms of safety, reliability, and robustness.

KAIST Multi-Spectral Day/Night Data Set for Autonomous and Assisted Driving

We introduce the KAIST multi-spectral data set, which covers a great range of drivable regions, from urban to residential, for autonomous systems. Our data set provides the different perspectives of the world captured in coarse time slots (day and night), in addition to fine time slots (sunrise, morning, afternoon, sunset, night, and dawn). For all-day perception of autonomous systems, we propose the use of a different spectral sensor, i.e., a thermal imaging camera. Toward this goal, we develop a multi-sensor platform, which supports the use of a co-aligned RGB/Thermal camera, RGB stereo, 3-D LiDAR, and inertial sensors (GPS/IMU) and a related calibration technique. We design a wide range of visual perception tasks including the object detection, drivable region detection, localization, image enhancement, depth estimation, and colorization using a single/multi-spectral approach. In this paper, we provide a description of our benchmark with the recording platform, data format, development toolkits, and lessons about the progress of capturing data sets.

Risk assessment for automatic lane change maneuvers on highways

This paper presents a risk assessment algorithm for automatic lane change maneuvers on highways. It is capable of reliably assessing a given highway situation in terms of the possibility of collisions and robustly giving a recommendation for lane changes. The algorithm infers potential collision risks of observed vehicles based on Bayesian networks considering uncertainties of its input data. It utilizes two complementary risk metrics (time-to-collision and minimal safety margin) in temporal and spatial aspects to cover all risky situations that can occur for lane changes. In addition, it provides a robust recommendation for lane changes by filtering out uncertain noise data pertaining to vehicle tracking. The validity of the algorithm is tested and evaluated on public highways in real traffic as well as a closed high-speed test track in simulated traffic through in-vehicle testing based on overtaking and overtaken scenarios in order to demonstrate the feasibility of the risk assessment for automatic lane change maneuvers on highways.

A unified index for moving-objects databases

Recently, the need for Location-Based Services (LBS) has increased due to the development and widespread use of mobile devices (e.g., PDAs, cellular phones, laptop computers, GPS, and RFID etc). The core technology of LBS is a moving-objects database that stores and manages the positions of moving objects. To search for information quickly, the database needs to contain an index that supports both real-time position tracking and management of large numbers of updates. As a result, the index requires a structure operating in the main memory for real-time processing and requires a technique to migrate part of the index from the main memory to disk storage (or from disk storage to the main memory) to manage large volumes of data. To satisfy these requirements, this paper suggests a unified index scheme unifying the main memory and the disk as well as migration policies for migrating part of the index from the memory to the disk during a restriction in memory space. Migration policy determines a group of nodes, called the migration subtree, and migrates the group as a unit to reduce disk I/O. This method takes advantage of bulk operations and dynamic clustering. The unified index is created by applying various migration policies. This paper measures and compares the performance of the migration policies using experimental evaluation.

Toward highly automated driving by vehicle-to-infrastructure communications

This paper presents a cooperative system by vehicle-to-infrastructure (V2I) communications that extends the range of environmental perception and improves the performance of situation awareness for highly automated driving. The paper consists of two steps: data fusion based situation awareness and distributed reasoning based situation assessment. The data fusion produces a V2I augmented map to provide a better understanding of driving situations by integrating road infrastructures with a high-precision map. The distributed reasoning evaluates a risky level of a current situation in terms of road infrastructures through the use of independent local experts which are distributed into lane-level local regions of the vehicles surroundings. The recommendations for driving behaviors are determined by the combination of results from each expert. The system is tested and evaluated through in-vehicle tests on a highway test road to verify that it can determine appropriate reactions under road hazard situations, such as black ice and construction.