Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization
An autonomous car must know where it is with high precision in order to maneuver safely and reliably in both urban and highway environments. Thus, in this paper, a reliable and relatively precise position estimation (localization) technique for autonomous vehicles is proposed and implemented. In dea...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-12-01
|
| Series: | World Electric Vehicle Journal |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2032-6653/15/1/5 |
| _version_ | 1797339267353018368 |
|---|---|
| author | Wael A. Farag Julien Moussa H. Barakat |
| author_facet | Wael A. Farag Julien Moussa H. Barakat |
| author_sort | Wael A. Farag |
| collection | DOAJ |
| description | An autonomous car must know where it is with high precision in order to maneuver safely and reliably in both urban and highway environments. Thus, in this paper, a reliable and relatively precise position estimation (localization) technique for autonomous vehicles is proposed and implemented. In dealing with the obtained sensory data or given knowledge about the vehicle’s surroundings, the proposed method takes a probabilistic approach. In this approach, the involved probability densities are expressed by keeping a collection of samples selected at random from them (Monte Carlo simulation). Consequently, this Monte Carlo sampling allows the resultant position estimates to be represented with any arbitrary distribution, not only a Gaussian one. The selected technique to implement this Monte-Carlo-based localization is Bayesian filtering with particle-based density representations (i.e., particle filters). The employed particle filter receives the surrounding object ranges from a carefully tuned Unscented Kalman Filter (UKF) that is used to fuse radar and lidar sensory readings. The sensory readings are used to detect pole-like static objects in the egocar’s surroundings and compare them to the ones that exist in a supplied detailed reference map that contains pole-like landmarks that are produced offline and extracted from a 3D lidar scan. Comprehensive simulation tests were conducted to evaluate the outcome of the proposed technique in both lateral and longitudinal localization. The results show that the proposed technique outperforms the other techniques in terms of smaller lateral and longitudinal mean position errors. |
| first_indexed | 2024-03-08T09:44:28Z |
| format | Article |
| id | doaj.art-3a7dd1d0392a45b393f39a8391df99c4 |
| institution | Directory Open Access Journal |
| issn | 2032-6653 |
| language | English |
| last_indexed | 2024-03-08T09:44:28Z |
| publishDate | 2023-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | World Electric Vehicle Journal |
| spelling | doaj.art-3a7dd1d0392a45b393f39a8391df99c42024-01-29T14:26:19ZengMDPI AGWorld Electric Vehicle Journal2032-66532023-12-01151510.3390/wevj15010005Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo LocalizationWael A. Farag0Julien Moussa H. Barakat1College of Engineering and Technology, American University of the Middle East, Egaila 54200, KuwaitCollege of Engineering and Technology, American University of the Middle East, Egaila 54200, KuwaitAn autonomous car must know where it is with high precision in order to maneuver safely and reliably in both urban and highway environments. Thus, in this paper, a reliable and relatively precise position estimation (localization) technique for autonomous vehicles is proposed and implemented. In dealing with the obtained sensory data or given knowledge about the vehicle’s surroundings, the proposed method takes a probabilistic approach. In this approach, the involved probability densities are expressed by keeping a collection of samples selected at random from them (Monte Carlo simulation). Consequently, this Monte Carlo sampling allows the resultant position estimates to be represented with any arbitrary distribution, not only a Gaussian one. The selected technique to implement this Monte-Carlo-based localization is Bayesian filtering with particle-based density representations (i.e., particle filters). The employed particle filter receives the surrounding object ranges from a carefully tuned Unscented Kalman Filter (UKF) that is used to fuse radar and lidar sensory readings. The sensory readings are used to detect pole-like static objects in the egocar’s surroundings and compare them to the ones that exist in a supplied detailed reference map that contains pole-like landmarks that are produced offline and extracted from a 3D lidar scan. Comprehensive simulation tests were conducted to evaluate the outcome of the proposed technique in both lateral and longitudinal localization. The results show that the proposed technique outperforms the other techniques in terms of smaller lateral and longitudinal mean position errors.https://www.mdpi.com/2032-6653/15/1/5UKFADASautonomous drivingparticle filterMonte Carlolocalization |
| spellingShingle | Wael A. Farag Julien Moussa H. Barakat Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization World Electric Vehicle Journal UKF ADAS autonomous driving particle filter Monte Carlo localization |
| title | Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization |
| title_full | Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization |
| title_fullStr | Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization |
| title_full_unstemmed | Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization |
| title_short | Utilizing Probabilistic Maps and Unscented-Kalman-Filtering-Based Sensor Fusion for Real-Time Monte Carlo Localization |
| title_sort | utilizing probabilistic maps and unscented kalman filtering based sensor fusion for real time monte carlo localization |
| topic | UKF ADAS autonomous driving particle filter Monte Carlo localization |
| url | https://www.mdpi.com/2032-6653/15/1/5 |
| work_keys_str_mv | AT waelafarag utilizingprobabilisticmapsandunscentedkalmanfilteringbasedsensorfusionforrealtimemontecarlolocalization AT julienmoussahbarakat utilizingprobabilisticmapsandunscentedkalmanfilteringbasedsensorfusionforrealtimemontecarlolocalization |