arXiv subjects

Sergio M. Savaresi

Publications and source records attributed to Sergio M. Savaresi.

11 recordsLinked to original sources

arXiv2026

LCF3D: A Robust and Real-Time Late-Cascade Fusion Framework for 3D Object Detection in Autonomous Driving

Accurately localizing 3D objects like pedestrians, cyclists, and other vehicles is essential in Autonomous Driving. To ensure high detection performance, Autonomous Vehicles complement RGB cameras with LiDAR sensors, but effectively combining these data sources for 3D object detection remains challenging. We propose LCF3D, a novel sensor fusion framework that combines a 2D object detector on RGB images with a 3D object detector on LiDAR point clouds. By leveraging multimodal fusion principles, we compensate for inaccuracies in the LiDAR object detection network. Our solution combines two key principles: (i) late fusion, to reduce LiDAR False Positives by matching LiDAR 3D detections with RGB 2D detections and filtering out unmatched LiDAR detections; and (ii) cascade fusion, to recover missed objects from LiDAR by generating new 3D frustum proposals corresponding to unmatched RGB detections. Experiments show that LCF3D is beneficial for domain generalization, as it turns out to be successful in handling different sensor configurations between training and testing domains. LCF3D achieves significant improvements over LiDAR-based methods, particularly for challenging categories like pedestrians and cyclists in the KITTI dataset, as well as motorcycles and bicycles in nuScenes. Code can be downloaded from: https://github.com/CarloSgaravatti/LCF3D.

cs.CV↗

arXiv2025

A Multimodal Hybrid Late-Cascade Fusion Network for Enhanced 3D Object Detection

We present a new way to detect 3D objects from multimodal inputs, leveraging both LiDAR and RGB cameras in a hybrid late-cascade scheme, that combines an RGB detection network and a 3D LiDAR detector. We exploit late fusion principles to reduce LiDAR False Positives, matching LiDAR detections with RGB ones by projecting the LiDAR bounding boxes on the image. We rely on cascade fusion principles to recover LiDAR False Negatives leveraging epipolar constraints and frustums generated by RGB detections of separate views. Our solution can be plugged on top of any underlying single-modal detectors, enabling a flexible training process that can take advantage of pre-trained LiDAR and RGB detectors, or train the two branches separately. We evaluate our results on the KITTI object detection benchmark, showing significant performance improvements, especially for the detection of Pedestrians and Cyclists.

cs.CV↗

arXiv2024

Semi-automated transmission control for motorcycle gearshift: design, data-driven tuning and experimental validation

This brief addresses the gearshifting problem for Semi-Automated Manual Transmissions (S-AMT) in powered two-wheelers, a powertrain setup that allows fast and smooth gear shifts with minimal modifications to the traditional manual powertrain layout. We show that with a proper synchronization between the electronic clutch and engine torque, excellent gearshift performance can be obtained, but requires precise parameters calibration. We thus propose the use of a data-driven approach with Constrained Bayesian Optimization to optimize control parameters. The procedure's effectiveness is demonstrated on a real vehicle, assessing performance in terms of optimality, convergence rate, and repeatability.

eess.SY↗

arXiv2023

An Add-on Model Predictive Control Strategy for the Energy Management of Hybrid Electric Tractors

The hybridization process has recently touched also the world of agricultural vehicles. Within this context, we develop an Energy Management Strategy (EMS) aiming at optimizing fuel consumption, while maintaining the battery state of charge. A typical feature of agricultural machines is that their internal combustion engine is speed controlled, tracking the reference requested by the driver. In view of avoiding any modification on this original control loop, an add-on EMS strategy is proposed. In particular, we employ a multi-objective Model Predictive Control (MPC), taking into account the fuel consumption minimization and the speed tracking requirement, including the engine speed controller in the predictive model. The proposed MPC is tested in an experimentally-validated simulation environment, representative of an orchard vineyard tractor.

eess.SY↗

arXiv2023

Non-Invasive Experimental Identification of a Single Particle Model for LiFePO4 Cells

The rapid spread of Lithium-ions batteries (LiBs) for electric vehicles calls for the development of accurate physical models for Battery Management Systems (BMSs). In this work, the electrochemical Single Particle Model (SPM) for a high-power LiFePO4 cell is experimentally identified through a set of non-invasive tests (based on voltage-current measurements only). The SPM is identified through a two-step procedure in which the equilibrium potentials and the kinetics parameters are characterized sequentially. The proposed identification procedure is specifically tuned for LiFePO4 chemistry, which is particularly challenging to model due to the non-linearity of its open circuit voltage (OCV) characteristic. The identified SPM is compared with a second-order Equivalent Circuit Model (ECM) with State of Charge dependency. Models performance is compared on dynamic current profiles. They exhibit similar performance when discharge currents peak up to 1C (RMSE between simulation and measures smaller than 20 mV) while, increasing the discharge peaks up to 3C, ECM's performance significantly deteriorates while SPM maintains acceptable RMSE (< 50 mV).

eess.SY↗

arXiv2023

Virtual-bike emulation in a series-parallel human-powered electric bike

Combining the advantages of standard bicycles and electrified vehicles, electric bikes (e-Bikes) are promising vehicles to reduce emission and traffic. The current literature on e-Bikes ranges from works on the energy management to the vehicle control to properly govern the human-vehicle interaction. This last point is fundamental in chain-less series bikes, where the link between the human and the vehicle behavior is only given by a control law. In this work, we address this problem in a series-parallel bike. In particular, we provide an extension of the virtual-chain concept, born for series bikes, and then we improve it developing a virtual-bike framework. Experimental results are used to validate the effectiveness of the solutions, when the cyclist is actually riding the bike.

eess.SY↗

arXiv2023

Twin-in-the-loop state estimation for vehicle dynamics control: theory and experiments

In vehicle dynamics control, many variables of interest cannot be directly measured, as sensors might be costly, fragile, or even not available. Therefore, real-time estimation techniques need to be used. The previous approach suffers from two main drawbacks: (i) the approximations due to model mismatch might jeopardize the performance of the final estimation-based control; (ii) each new estimator requires the calibration from scratch of a dedicated model. In this paper, we propose a twin-in-the-loop scheme, where the ad-hoc model is replaced by an accurate multibody simulator of the vehicle, typically available to vehicles manufacturers and suitable for the estimation of any onboard variable, coupled with a compensator within a closed-loop observer scheme. Given the black-box nature of the digital twin, a data-driven methodology for observer tuning is developed, based on Bayesian optimization. The effectiveness of the proposed estimation method for the estimation of vehicle states and forces, as compared to traditional model-based Kalman filtering, is experimentally shown on a dataset collected with a sports car.

eess.SY↗

arXiv2020

Experimental Automatic Calibration of a Semi-Active Suspension Controller via Bayesian Optimization

The End-of-Line (EoL) calibration of semi-active suspension systems for road vehicles is usually a critical and expensive task, needing a team of vehicle and control experts as well as many hours of professional driving. In this paper, we propose a purely data-based tuning method enabling the automatic calibration of the parameters of a proprietary suspension controller by relying on little experimental time and exploiting Bayesian Optimization tools. A detailed methodology on how to select the most critical degrees of freedom of the algorithm is also provided. The effectiveness of the proposed approach is assessed on a commercial multi-body simulator as well as on a real car.

eess.SY↗

arXiv2016

Robust linear static anti-windup with probabilistic certificates

In this paper, we address robust static anti-windup compensator design and performance analysis for saturated linear closed loops in the presence of nonlinear probabilistic parameter uncertainties via randomized techniques. The proposed static anti-windup analysis and robust performance synthesis correspond to several optimization goals, ranging from minimization of the nonlinear input/output gain to maximization of the stability region or maximization of the domain of attraction. We also introduce a novel paradigm accounting for uncertainties in the energy of the disturbance inputs. Due to the special structure of linear static anti-windup design, wherein the design variables are decoupled from the Lyapunov certificates, we introduce a significant extension, called scenario with certificates (SwC), of the so-called scenario approach for uncertain optimization problems. This extension is of independent interest for similar robust synthesis problems involving parameter-dependent Lyapunov functions. We demonstrate that the scenario with certificates robust design formulation is appealing because it provides a way to implicitly design the parameter-dependent Lyapunov functions and to remove restrictive assumptions about convexity with respect to the uncertain parameters. Subsequently, to reduce the computational cost, we present a sequential randomized algorithm for iteratively solving this problem. The obtained results are illustrated by numerical examples.

eess.SY↗

arXiv2015

Least costly energy management for series hybrid electric vehicles

Energy management of plug-in Hybrid Electric Vehicles (HEVs) has different challenges from non-plug-in HEVs, due to bigger batteries and grid recharging. Instead of tackling it to pursue energetic efficiency, an approach minimizing the driving cost incurred by the user - the combined costs of fuel, grid energy and battery degradation - is here proposed. A real-time approximation of the resulting optimal policy is then provided, as well as some analytic insight into its dependence on the system parameters. The advantages of the proposed formulation and the effectiveness of the real-time strategy are shown by means of a thorough simulation campaign.

eess.SY↗

arXiv2014

Scenario optimization with certificates and applications to anti-windup design

In this paper, we introduce a significant extension, called scenario with certificates (SwC), of the so-called scenario approach for uncertain optimization problems. This extension is motivated by the observation that in many control problems only some of the optimization variables are used in the design phase, while the other variables play the role of certificates. Examples are all those control problems that can be reformulated in terms of linear matrix inequalities involving parameter-dependent Lyapunov functions. These control problems include static anti-windup compensator design for uncertain linear systems with input saturation, where the goal is the minimization of the nonlinear gain from an exogenous input to a performance output. The main contribution of this paper is to show that randomization is a useful tool, specifically for anti-windup design, to make the overall approach less conservative compared to its robust counterpart. In particular, we demonstrate that the scenario with certificates reformulation is appealing because it provides a way to implicitly design the parameter-dependent Lyapunov functions. Finally, to further reduce the computational cost of this one-shot approach, we present a sequential randomized algorithm for iteratively solving this problem.

eess.SY↗