PRIDE: A Framework for Performance Evaluation of Intelligent Vehicles in Dynamic, On-Road Environments

We are developing a novel framework, PRIDE (PRediction In Dynamic Environments), to perform moving object prediction for unmanned ground vehicles. The underlying concept is based upon a multi-resolutional, hierarchical approach that incorporates multiple prediction algorithms into a single, unifying framework. The lower levels of the framework utilize estimation-theoretic short-term predictions while the upper levels utilize a probabilistic prediction approach based on situation recognition with an underlying cost model. In addition to predicting the location of moving objects in the environment, we have extended PRIDE to generate simulated traffic flow during on-road driving. In this paper, we explore applying the PRIDE-based traffic control algorithms for the purpose of performance evaluation of autonomous vehicles. Through the use of repeatable and realistic traffic flow simulation, one is able to evaluate the performance of an autonomous vehicle in an on-road driving scenario without the risk involved with introducing the vehicle into a potentially dangerous roadway situation. In addition, by varying a single vehicle's parameters (e.g. aggressivity, speed, location) with the traffic flow, we can show how the entire traffic pattern is affected. We will show the successes that have been achieved to date in a simulated environment, as well as enhancements that are currently being researched and expected in the near future.