|M.Sc Student||Amit Serebrennikov|
|Subject||Navigation and Guidance of a Man-Tracking Vehicle|
|Department||Department of Electrical Engineering||Supervisors||Full Professor Shimkin Nahum|
|Professor Emeritus Gutman Per-Olof|
We consider the problem of navigation and guidance of a human-tracking autonomous vehicle (AV). The use of the human leader is one of possible auxiliary means for off road AV navigation applications.
In contrast to standard guidance schemes, the vehicle must follow the precise path outlined by the human, at a pre-defined distance behind (of around 20 meters). This tracking mode allows the leader to choose a safe path for the follower, while keeping a safety distance between them. The main navigation challenge is to maintain an estimate of the leader's position for the last 20 meters of its path, and in particular to minimize the relative error between the vehicle's position and the required point on that path.
The relative position to the leader's current position is measured using a radar-like sensor. To know how the human was located in the past relative to AV at present, it is necessary to calculate previous car positions. We chose odometry only measurements as widely used and inexpensive method for AV localization. The drawback of odometry is integrating of incremental position errors, which causes to growth of absolute position error with time.
This paper presents an algorithm, which ensures a bounded error in the human’s path estimation relative to current AV position despite of odometry sensor errors. This algorithm is based on using the car incremental position estimates instead of its absolute position. We start by developing appropriate models for the human and vehicle motion, taking wheel slippage into account. We then present the tracking algorithm, which operates in the vehicle's coordinate frame, and incorporates Kalman filtering and smoothing. In addition, a literature survey on relevant issues in radar tracking and wheel slippage is presented.