George Rosario Jagadeesh

Heuristic techniques for accelerating hierarchical routing on road networks

The route computation module is one of the most important functional blocks in a dynamic route guidance system. Although various algorithms exist for finding the shortest path, their performance tends to deteriorate as the network size increases. We present an efficient hierarchical routing algorithm that finds a near-optimal route and evaluate it on a large city road network. Solutions provided by the hierarchical routing algorithm are compared with the optimal solutions to analyze and quantify the loss of accuracy. We propose a novel yet simple heuristic to substantially improve the performance of the hierarchical routing algorithm with acceptable loss of accuracy. A network pruning technique has been incorporated into the algorithm to reduce the search space and the correctness of the results is evaluated. The improved hierarchical routing algorithm that incorporates the heuristic techniques has been found to be over 50 times faster than a typical shortest path algorithm.

A map matching method for GPS based real-time vehicle location

Accurate vehicle location is essential for various applications in the field of intelligent transportation systems (ITS). Existing vehicle location systems rely on multiple positioning sensors and powerful computing devices to execute complex map matching algorithms. There exists a strong need for exploring a solution for vehicle location that relies on a GPS receiver as the sole means of positioning and does not require complex computations. Towards this end, the error characteristics of the GPS signal were studied through the analysis of GPS data collected during test drives. Based on the inferences drawn and a simple fuzzy rule set, a novel yet simple map matching algorithm was developed. Due to the difficulties in testing the algorithm through on-road trials, a simulation environment that is capable of reproducing the field conditions in the laboratory was developed. Simulation results confirm that the proposed algorithm overcomes many of the inadequacies of the existing methods and is capable of achieving high accuracy with minimal computational requirements.

Online Map-Matching of Noisy and Sparse Location Data With Hidden Markov and Route Choice Models

With the growing use of crowdsourced location data from smartphones for transportation applications, the task of map-matching raw location sequence data to travel paths in the road network becomes more important. High-frequency sampling of smartphone locations using accurate but power-hungry positioning technologies is not practically feasible as it consumes an undue amount of the smartphone’s bandwidth and battery power. Hence, there exists a need to develop robust algorithms for map-matching inaccurate and sparse location data in an accurate and timely manner. This paper addresses the above-mentioned need by presenting a novel map-matching solution that combines the widely used approach based on a hidden Markov model (HMM) with the concept of drivers’ route choice. Our algorithm uses an HMM tailored for noisy and sparse data to generate partial map-matched paths in an online manner. We use a route choice model, estimated from real drive data, to reassess each HMM-generated partial path along with a set of feasible alternative paths. We evaluated the proposed algorithm with real world as well as synthetic location data under varying levels of measurement noise and temporal sparsity. The results show that the map-matching accuracy of our algorithm is significantly higher than that of the state of the art, especially at high levels of noise.

Probabilistic Map Matching of Sparse and Noisy Smartphone Location Data

There is an immense amount of location data being collected today from smartphone users by various service providers. Due to bandwidth and battery-life considerations, smartphone locations are generally sampled at sparse intervals using energy-efficient, but inaccurate, alternatives to the power-hungry Global Positioning System (GPS). If sparse sequences of coarse location data obtained from mobile users can be accurately map-matched to travel paths on the road network, then this data can be effectively used for several traffic-related applications. Unlike most other map-matching methods in the literature, we, in this paper, focus on the problem of map-matching sparse and noisy non-GPS smartphone location data. We adopt the widely-followed Hidden Markov Model (HMM) approach and propose new probabilistic models for the observation and transition probabilities tailored towards the type of data being considered. Our map-matching method has been evaluated using ground-truth labelled non-GPS location data collected from real drives. Tests show that the accuracy of the proposed method is about 12% more than that of a comparable HMM-based method from the literature. Our results also show that the runtime and latency of the proposed method can be kept within reasonable bounds using simple techniques.

Robust real-time route inference from sparse vehicle position data

The ability to correctly infer the route traveled by vehicles in real time from infrequent, noisy observations of their position is useful for several traffic management applications. This task, known as map matching, is efficiently performed through probabilistic inference on a Hidden Markov Model that represents the candidate vehicle states and the transitions between them. In this paper, we present new methods for improving the accuracy and timeliness of existing solutions. We propose assigning the transition probability between a pair of candidate vehicle states by considering the alternative paths present in the context. A discrete route choice model is used to estimate the probability that a driver would choose the path under consideration over the best alternative available. In order to facilitate real-time operation, we present a simple yet effective heuristic to reduce the output latency of the route-inference algorithm with negligible loss of accuracy. Tests conducted with ground truth GPS data from a dense urban region in Singapore show that the proposed techniques outperform the conventional baseline approach.

Notice of Retraction Practical algorithm for shortest path on large networks with time-dependent edge-length

The shortest path problem in network has been studied widely and intensively over years, both in theoretical and in computational viewpoints. Many speed-up techniques for Dijkstras algorithm have been developed. However, only few of those techniques work in time-dependent networks. This paper studies how to promptly answer the shortest path between a pair of nodes over a large time-dependent road network, to cater for the increasing interest in the advanced transportation systems. A fast algorithm is proposed for an approximate time-dependent shortest path (TDSP). The proposed algorithm works on a subnetwork derived from the static shortest path between the given pair of nodes. The size of the subnetwork can be modulated according to the given traffic period for high solution quality. An extensive experimental results on a real-world environments (Singapore road network) show that the traditional algorithm is improved by up to 90% in terms of runtime. In addition, about 65% random instances can be exactly solved and most instances can be solved within the error of 3%.

A clustering-based approach for data-driven imputation of missing traffic data

The problem of missing samples in road traffic data undermines the performance of intelligent transportation applications. This paper proposes a data-driven imputation method that exploits the spatial and temporal relationships existing between the traffic flows of multiple road segments that are correlated with each other. The K-means clustering technique is used to group together road segments with similar traffic flow patterns. Next, a deep-learning model based on stacked denoising autoencoders is constructed for each group of road segments to extract their spatial-temporal relationships and use them for imputing the missing data points. Experiments conducted with real traffic data demonstrate that the imputation accuracy of the proposed method is robust under different missing data rates.

Heuristic optimizations for high-speed low-latency online map matching with probabilistic sequence models

The computation speed and output latency of map matching are important considerations when processing location data, especially smartphone-generated noisy and sparse data, from a large number of users for real-time transportation applications. In this paper, we examine the factors affecting the efficiency of online map matching algorithms that are based on probabilistic sequence models such as Hidden Markov Models (HMM) and present several heuristic optimizations to improve their speed and latency. As shortest path computations account for most of the running time of probabilistic map matching algorithms, we propose a method for reducing the total number of such computations by pruning unlikely states in the probabilistic sequence model. Furthermore, we speed up the one-to-many shortest path computations by limiting the search space to an elliptical area that encompasses all the targeted destinations. We present a technique for reducing the latency of the Viterbi algorithm used to find the most likely state sequence in a HMM or a similar model. This technique enables the early output of partial state sequences based on an estimate of the probability of a state being part of the eventual most likely sequence. Experiments using real-world location data show that the heuristic optimizations significantly reduce the running time and output latency with negligible loss of accuracy.

A Short Course on Implementing FPGA Based Digital Systems

The rapid advances in the FPGA technology along with high-levels of system integration have made FPGAs the preferred platform not only for rapid prototyping but also for production of digital embedded systems. This paper presents the experience of a team of instructors in designing and conducting a short course on implementing FPGA-based digital systems for industry professionals. The selection of topics, course organization, the issues involved in designing effective hands-on exercises and the response of the students to the course are discussed.

A Low-Complexity Speaker-and-Word Recognition Application for Resource-Constrained Devices

We present a low-complexity solution for performing speaker-and-word recognition and demonstrate its suitability for resource-constrained embedded / mobile devices. In the proposed approach, modeling and recognition of speakers and words are performed using Gaussian Mixture Model (GMM), which has relatively low computational complexity. The inability of GMM to capture the temporal information of speech, which is vital for word recognition, has been overcome through a simple, yet effective adaptation. After evaluating the performance of two alternative architectures, an integrated speaker-and-word recognition system based on text-dependent speaker recognition has been proposed. The system has been ported to a mobile device as an Android application and tested in real-life environment.