arXiv subjects

Chouaib Bencheikh Lehocine

Publications and source records attributed to Chouaib Bencheikh Lehocine.

4 recordsLinked to original sources

arXiv2026

Beyond Chamfer Distance: Granular Order-aware Evaluation Metric For Online Mapping

Online map estimation is a crucial component of autonomous driving systems that reduces the reliance on costly high-definition maps. State-of-the-art (SOTA) methods commonly predict map elements as ordered sequences of points that form polylines and polygons. The evaluation of these methods relies predominantly on mean average precision (mAP) based on thresholded Chamfer distance (CD). This framework lacks sensitivity to point ordering and provides limited granularity in assessing geometric quality, making it difficult to distinguish which methods truly excel over others. In this work, we address these limitations on two fronts. For the single-instance similarity measure, we introduce sequence optimal sub-pattern assignment (SOSPA), an order-aware metric that enables fine-grained evaluation of individual geometries while satisfying all metric axioms. For the multi-instance evaluation framework, we propose polyline localisation and detection (PLD), a soft metric that jointly captures detection quality and geometric accuracy, replacing the hard thresholding of mAP with a principled soft assignment. Through evaluations on nuScenes, we demonstrate that PLD effectively ranks SOTA online mapping methods (MapTRv2, StreamMapNet, MapTracker) while providing a decomposed error analysis. This analysis identifies detection capability as the dominant bottleneck in current methods, revealing a performance trend that mAP fails to capture. Code for evaluation using our metrics will be released.

cs.CV↗

arXiv2022

Antenna Combiner for Periodic Broadcast V2V Communication Under Relaxed Worst-Case Propagation

The performance of a previously developed analog combining network (ACN) of phase shifters for vehicle-to-vehicle communication is investigated. The original ACN was designed to maximize the sum of the signal-to-noise ratios (SNRs) for $K$ consecutive, broadcast, periodic cooperative awareness messages when communication is over a dominant path whose angle of arrival (AOA) is constant over the duration of $K$ packets. In this work, we relax this scenario by allowing the dominant path AOA and path-loss (PL) to be time-variant. Assuming a highway scenario with line of sight (LOS) propagation between vehicles, we use affine approximations to model the time variation of different path quantities, including the PL, the relative distance-dependent phase shift between antennas, and the AOA-dependent far-field function of the antennas. Using these approximations, we analytically derive the ACN sum-SNRs as each one of these quantities vary over the duration of $K$ packets. Moreover, we suggest a phase slope design rule that is robust against time variation of the dominant path and optimal under time-invariant conditions. Finally, we validate this design rule using numerical computations and an example of vehicular communication antenna elements.

cs.IT↗

arXiv2021

Robust Analog Beamforming for Periodic Broadcast V2V Communication

We generalize an existing low-cost analog signal processing concept that takes advantage of the periodicity of vehicle-to-vehicle broadcast service to the transmitter side. In particular, we propose to process multiple antennas using either an analog beamforming network (ABN) of phase shifters, or an an antenna switching network (ASN) that periodically alternate between the available antennas, to transmit periodic messages to receivers that have an analog combining network (ACN) of phase shifters, which has been proposed in earlier work. To guarantee robustness, we aim to minimize the burst error probability for the worst receiving vehicular user, in a scenario of bad propagation condition that is modeled by a single dominant path between the communicating vehicles. In absence of any form of channel knowledge, we analytically derive the optimal parameters of both ABN and ASN. The ABN beamforming vector is found to be optimal for all users and not only for the worst receiving user. Further, we demonstrate that Alamouti scheme for the special case of two transmit antennas yields similar performance to ABN and ASN. At last, we show that the derived parameters of the two proposed transmission strategies are also optimal when hybrid ACN-maximal ratio combining is used at the receiver.

cs.IT↗

arXiv2021

Hybrid Combining of Directional Antennas for Periodic Broadcast V2V Communication

A hybrid analog-digital combiner for broadcast vehicular communication is proposed. It has an analog part that does not require any channel state information or feedback from the receiver, and a digital part that uses maximal ratio combining (MRC). We focus on designing the analog part of the combiner to optimize the received signal strength along all azimuth angles for robust periodic vehicle-to-vehicle (V2V) communication, in a scenario of one dominant component between the communicating vehicles (e.g., highway scenario). We show that the parameters of a previously suggested fully analog combiner solves the optimization problem of the analog part of the proposed hybrid combiner. Assuming $L$ directional antennas with uniform angular separation together with the special case of a two-port receiver, we show that it is optimal to combine groups of $\lceil L/2\rceil$ and $\lfloor L/2\rfloor$ antennas in analog domain and feed the output of each group to one digital port. This is shown to be optimal under a sufficient condition on the sidelobes level of the directional antenna. Moreover, we derive a performance bound for the hybrid combiner to guide the choice of antennas needed to meet the reliability requirements of the V2V communication links.

cs.IT↗