A parallel implementation of a multisensor feature-based range-estimation method

Abstract

There are many proposed vision based methods to perform obstacle detection and avoidance for autonomous or semi-autonomous vehicles. All methods, however, will require very high processing rates to achieve real time performance. A system capable of supporting autonomous helicopter navigation will need to extract obstacle information from imagery at rates wrying from ten frames per second to thirty or more frames per second depending on the vehicle speed. Such a system will need to sustain billions of operations per second. To reach such high processing rates using current technology, a parallel implementation of the obstacle detection/ranging method is required. This paper describes an efficient and flexible parallel implementation of a multisensor feature-based range-estimation algorithm, targeted for helicopter flight, realized on both a distributed-memory and shared-memory parallel computer. 17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION OF REPORT OF THIS PAGE

State Estimation for Autonomous Helicopter via Sensor Modeling

“…In [6], the authors have developed a vision based system that allows an autonomous helicopter to perform a line tracking task. Vision based obstacle avoidance was described in [7] where a multi-sensor feature-based range-estimation algorithm was proposed for automated helicopter flight. This algorithm could track many features at the same time in multiple image sensors using an extended Kalman filter to estimate the feature locations in a master sensor coordinate frame.…”

Accuracy bounds and optimal computation of robot localization

“…Here we consider the case in which a robot already has a three-dimensional (3-D) map of the environment and study the problem of identifying its current location relative to the world model. In theory, the current location can be computed by tracing the history of motion from a known initial position, e.g., by integrating the rotation of the wheels or incrementally correcting the position (Suorsa and Sridhar 1994). However, the accuracy of the computed location quickly deteriorates as errors (due to slippage of the wheels, vibrations of the camera, etc.)…”

<title>Real-time computational needs of a multisensor feature-based range estimation method</title>

“…This paper is a continuation of a previous work which described, at length, a parallel version of our feature-based range-estimation method (hereafter referred to as Opt ow) 7 ]. Within that work Opt ow was ported to a distributed computer (based on a network of eight w orkstations) and a multithreaded sharedmemory multicomputer (a Silicon Graphics IRIS 4D/480).…”

“…A coarse-grained thread approach w as chosen to parallelize Opt ow on this architecture. 7 In this approach N threads are generated for an N-processor machine. The number of VPRs should exceed the numb e r o f p r o c e s s o r s s u c h that the load balancer will have enough resolution to approach t h e T= N lower bound.…”