Bernd Lichte

Multi-Target Tracking using a 3D-Lidar sensor for autonomous vehicles

Environmental perception is a prerequisite for autonomous driving and also a challenging task particularly in cluttered dynamic environments such as complex urban situations. In this paper, we present a robust algorithm for Multi-Target Tracking (MTT) using a Velodyne 3D HDL-64 Lidar sensor. The main contribution of this paper is a practical framework for selecting and representing useful information from the sensor raw data. Since the sensor produces a huge amount of data, a perception algorithm cannot be carried out in real-time without simplifying the sensor information. Unlike prior works, we introduce hybrid ground classification and the Region of Interest (ROI) identification method in order to filter out the amount of unwanted raw data for the actual tracking. And the environment is also abstracted based on an occupancy grid map. Moreover, we introduce feature based object geometry for precise estimation of the system state. In contrast to prior approaches, which use object geometry for the classification, we use it in order to compensate the unintended dynamics caused by shape change or occlusion. Our proposed MTT algorithm is able to run in real-time with an average processing time of 20ms. We evaluate it using our experimental vehicle “Leonie” in complex urban scenarios.

A surveillance and safety system based on performance criteria and functional degradation for an autonomous vehicle

Autonomous driving in urban environments is potentially dangerous since a malfunction of vehicle guidance systems can lead to severe situations for passengers inside the autonomous vehicle and other road users. Therefore both, monitoring the current system operation state by a surveillance system, which is able to detect failures of software and hardware modules, and a safety system, which reacts on these failures immediately, is necessary. In this paper an approach based on performance criteria and functional degradation is proposed, which is used in the autonomous vehicle Leonie developed within the Stadtpilot project. The surveillance part of the system collects data from sensors, software modules, hardware, and the vehicle to combine this data with heuristics to performance criteria. Based on these criteria degradation actions are executed to keep the operation of Leonie in a safe state. The safety system can influence driving maneuvers like lane changes and turning maneuvers, modify driving parameters like maximum speed and safe time headway and even force driving maneuvers like emergency stops and controlled stops at the side of the road. Currently, the safety driver onboard of Leonie is the fallback solution in case of a system malfunction. Using the proposed safety system should reduce the number of situations where the safety driver has to take control over the vehicle though.

Safe, dynamic and comfortable longitudinal control for an autonomous vehicle

Driver assistance systems are commonly available in many vehicles. There are systems for safety functions like the Electronic Stability Control, Automatic Traction Control, Anti-lock Brake System and automatic emergency braking. There are also systems for comfort functions like adaptive cruise control with stop and go functionality and combined safety and comfort functions like lane keeping and side-wind assistance. A control system consisting of all of these systems would allow comfortable automatic vehicle guidance on highways. In an urban environment, like in the Stadtpilot project, requirements on driver assistance systems are higher, especially in the case of full autonomous driving. An essential part of an autonomous vehicle control system is a longitudinal controller for acceleration and deceleration of the vehicle. This longitudinal control system has to take care of many more conditions than an assistance system. E.g. it needs to perceive and calculate road and weather conditions with its sensors, which usually is a task a human driver does instinctively. The present paper describes how the autonomous vehicle Leonie is able to adapt its longitudinal control to changing road and weather conditions by calculating a so called Grip Value and gives an outlook how this parameter affects whole vehicle guidance.

Probabilistic fault detection and handling algorithm for testing stability control systems with a drive-by-wire vehicle

This paper presents a probabilistic fault detection and handling algorithm (PFDH) for redundant and deterministic X-by-wire systems. The algorithm is specifically designed to guarantee safe operation of an experimental drive-by-wire vehicle used as test platform and development tool in research projects focusing on vehicle dynamics. The required flexibility of the overall system for use as a test bed influences significantly the redundancy structure of the onboard network. A “black box” approach to integrate newly developed user algorithms is combined with a hot-standby architecture controlled by PFDH. This way, functional redundancy for basic driving operations can be achieved despite unknown software components. PFDH is based on monitoring multiple criteria over time, including vehicle dynamics and relative error probabilities of hard- and software components provided by experts or statistical data.

Consistency-based motion classification for laser sensors dealing with cross traffic in urban environments

In this paper an approach for a consistency-based motion classification for laser sensors is presented which concentrates on urban environments. In these complex environments the algorithm has to match both cross traffic and structures in parallel to the road, as well as objects starting and stopping moving. This leads to a conflict to be solved. For a better understanding we introduce some basic definitions at the beginning. As there are limits due to the sensors properties, the proposed algorithm can be configured. The parameters depend on the special dynamic characteristics of the scenario to be detected on the one hand and on the other hand on the sensors properties. In combination with the resulting speed limit of the ego vehicle, these parameters describe the theoretical limits of this approach in a comprehensible way. This approach runs online and has been validated in crowded urban environment.