Sensing and Control for Autonomous Vehicles by Thor I. Fossen Kristin Y. Pettersen & Henk Nijmeijer

Author:Thor I. Fossen, Kristin Y. Pettersen & Henk Nijmeijer
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

2 System Architecture

The architecture for the tracking system that will be described in this paper is shown in Fig. 1. Each UAS instantiates a tracking pipeline that includes six key components. The first component in the pipeline, as shown in Fig. 1, is the UAS and gimbaled camera. We assume that the UAS contains an autopilot system, as well as a pan-tilt camera that can be automatically controlled to point along a desired optical axis. In this paper, we will assume an RGB camera and enough processing power on-board to process images at frame rate, and to implement the other components in the system. The second component shown in Fig. 1 is the Geolocation block. The purpose of the geolocation block is to transform the image coordinates into world coordinates based on the current pose of the UAS. A detailed description of the geolocation block is given in Sect. 2.2. The next component shown in Fig. 1 is the Recursive RANdom SAmple Consensus Multiple Target Tracking (R-RANSAC MTT) block. This block uses image features in world coordinates to create and manage object tracks. This block performs several key tasks including data association , new track formation, track propagation, track collation, and track deletion. We define a track to be the time history of the system state (position, velocity, acceleration, etc.), as well as the associated covariance matrix. A more detailed description of this block will be given in Sect. 2.3. The current tracks maintained by each UAS is shared across the network with other UAS. The current collection of tracks is used in the Bias Estimation block shown in Fig. 1 to estimate the translational and rotational bias between each pair of tracks in the network, and thereby place all tracks in the coordinate system of the UAS. Additional details about this process will be described in Sect. 2.4. The collection of tracks are then processed by the Track-to-track Association block shown in Fig. 1. This block uses a statistical test on a past window of the data to determine which tracks maintained by the UAS are statistically similar to the tracks maintained by the UAS. The details of this block are described in Sect. 2.5. When tracks are determined to be similar, they are fused in the Track Fusion block shown in Fig. 1. Track fusion is accomplished using an information consensus filter, as described in Sect. 2.6.

Fig. 1Architecture for tracking multiple ground-based objects of interest using a team of UAS

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.